Release v2.0.0 — single-file .nbq models + first-run download (CPU + CUDA); promote from v2-preview

.gitattributes

@@ -61,3 +61,13 @@ MedpodInstaller-1.3.5-cpu-avx512.exe filter=lfs diff=lfs merge=lfs -text
 MedpodInstaller-1.3.6-cpu.exe filter=lfs diff=lfs merge=lfs -text
 MedpodInstaller-1.3.6-cpu-raptorlake.exe filter=lfs diff=lfs merge=lfs -text
 MedpodInstaller-1.3.6-cpu-avx512.exe filter=lfs diff=lfs merge=lfs -text
+MedPodGPInstaller-2.0.0-cpu.exe filter=lfs diff=lfs merge=lfs -text
+MedPodGPInstaller-2.0.0-cpu-raptorlake.exe filter=lfs diff=lfs merge=lfs -text
+MedPodGPInstaller-2.0.0-cpu-avx512.exe filter=lfs diff=lfs merge=lfs -text
+MedPodGPInstaller-2.0.0-cuda.exe filter=lfs diff=lfs merge=lfs -text
+medgemma-1.5-4b-text-q4km-cpu.nbq filter=lfs diff=lfs merge=lfs -text
+medgemma-1.5-4b-vision-cpu.nbq filter=lfs diff=lfs merge=lfs -text
+medgemma-1.5-4b-vision-cuda.nbq filter=lfs diff=lfs merge=lfs -text
+qwen3-asr-0.6b-int8-cpu.nbq filter=lfs diff=lfs merge=lfs -text
+qwen3-asr-0.6b-f16-cuda.nbq filter=lfs diff=lfs merge=lfs -text
+qwen3.5-0.8b-summariser-q4km-cpu.nbq filter=lfs diff=lfs merge=lfs -text

LICENSE.md

@@ -1,150 +1,150 @@
----
-title: "MedPod GP — End-User License Agreement"
-subtitle: "Free-Distribution EULA · proprietary clinical software, free of charge"
-issue: "1.0"
-date: "2026-06-27"
-classification: "Public"
----
-
-# MedPod GP — End-User License Agreement
-
-Copyright © 2026 CloudKites Pty Ltd (CloudKites AI Lab), ABN 91 633 195 874,
-Sydney, NSW, Australia. All rights reserved.
-
-MedPod GP is **free-of-charge** software distributed in compiled (binary) form only
-for Windows. MedPod GP and "Numbat" (the on-device machine-learning toolkit embedded
-in MedPod GP) are proprietary software and the exclusive intellectual property of
-CloudKites Pty Ltd ("CloudKites"). Their source code is confidential and is **NOT**
-distributed. This Agreement governs your use and redistribution of the MedPod GP
-software and its accompanying materials (the "Software").
-
-> **BY INSTALLING, USING, OR DISTRIBUTING THE SOFTWARE YOU AGREE TO THIS
-> AGREEMENT. IF YOU DO NOT AGREE, DO NOT INSTALL, USE, OR DISTRIBUTE THE
-> SOFTWARE.**
-
-## The Agreement
-
-### 1. License Grant (Free of Charge)
-
-CloudKites grants you a free, worldwide, non-exclusive, royalty-free, perpetual
-license to install and use the Software, in object-code (binary) form, on any
-number of devices you own or control, for personal or professional purposes, at no
-charge, subject to the intended-use and clinical-responsibility terms in Section 6.
-
-### 2. Redistribution (Permitted, Non-Commercial)
-
-You may copy and redistribute the Software free of charge, provided that:
-
-- **(a)** you distribute only the **complete, unmodified** official package or
-  installer, exactly as supplied by CloudKites;
-- **(b)** you distribute it **free of charge** and do not sell it, charge any fee
-  for it, place it behind a paywall, or accept any other consideration for it;
-- **(c)** you do not bundle, embed, or integrate it into any paid product, paid
-  service, or commercial offering; and
-- **(d)** all copyright, trademark, license, and attribution notices (including
-  this Agreement and `THIRD_PARTY_LICENSES.md`) remain intact and are included with
-  every copy.
-
-Any other distribution requires the prior written permission of CloudKites.
-
-### 3. Restrictions
-
-You may **NOT**, and may not permit any third party to:
-
-- **(a)** sell, rent, lease, sublicense, or charge any fee for the Software, or
-  distribute it other than as permitted in Section 2;
-- **(b)** modify, adapt, translate, repackage, or create derivative works of the
-  Software;
-- **(c)** reverse engineer, decompile, disassemble, or otherwise attempt to derive,
-  reconstruct, or extract the source code, model architectures, or trade secrets of
-  the Software or of Numbat, except to the extent this restriction is prohibited by
-  applicable law;
-- **(d)** remove, alter, or obscure any proprietary, copyright, trademark, license,
-  or attribution notices; or
-- **(e)** upload, post, or otherwise place the **source code** of the Software or of
-  Numbat on any public repository or public-access network.
-
-The Software's source code and Numbat's source code are strictly confidential and
-are **NOT** distributed under this Agreement.
-
-### 4. Ownership
-
-The Software and Numbat are licensed, not sold. CloudKites retains all right, title,
-and interest, including all intellectual-property rights. No rights are granted
-other than as expressly set out in this Agreement.
-
-### 5. Third-Party Components and AI Models
-
-The Software bundles or downloads third-party open-source libraries and third-party
-AI model weights, each provided under its own license. Those components are licensed
-to you under their respective terms (the Qt framework under LGPL-3.0; FFmpeg under
-LGPL-2.1-or-later; SQLite in the public domain; for GPU builds the NVIDIA CUDA and
-cuDNN runtimes under the NVIDIA Software License Agreement; the MedGemma / Gemma
-models under the Gemma Terms of Use; the Qwen models under Apache-2.0; and others),
-which prevail over this Agreement for those components only. Required attributions
-and license texts are provided in `THIRD_PARTY_LICENSES.md`, which forms part of this
-Agreement and must be retained in any redistribution. Your use of bundled AI models
-is additionally subject to those models' acceptable-use / prohibited-use policies
-(notably the Gemma Terms of Use and Prohibited Use Policy for the Gemma / MedGemma
-models).
-
-### 6. Medical / Clinical Use — No Reliance; Not a Medical Device
-
-MedPod GP is a **clinical documentation and decision-support aid for use by, or under
-the supervision of, a qualified healthcare professional.** It is **NOT a medical
-device**, is **not intended to diagnose, treat, cure, or prevent any disease**, and
-has **not** been cleared, approved, or certified by the TGA, FDA, MHRA, CE/EU MDR, or
-any other regulatory authority.
-
-The Software uses generative and predictive AI models (for transcription, chat,
-image analysis, search, and note drafting). Their output may be **inaccurate,
-incomplete, biased, or fabricated, even when it appears confident**, and may contain
-transcription or recognition errors. Output is **not** medical advice and is **not** a
-substitute for the independent professional judgement of a qualified clinician.
-
-You are solely responsible for any clinical decision, diagnosis, documentation, or
-action taken, and you must **independently verify all AI-generated content** —
-including transcriptions, summaries, image interpretations, search results, and
-drafted notes — against the patient, the source material, and trusted clinical
-references before relying on it or entering it into a medical record. You are
-responsible for compliance with all laws and professional obligations applicable to
-you, including patient privacy and health-records law.
-
-### 7. No Warranty
-
-THE SOFTWARE IS PROVIDED FREE OF CHARGE, "AS IS" AND "AS AVAILABLE", WITHOUT
-WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING WARRANTIES OF MERCHANTABILITY,
-FITNESS FOR A PARTICULAR PURPOSE, ACCURACY, CLINICAL ACCURACY, AND NON-INFRINGEMENT,
-TO THE MAXIMUM EXTENT PERMITTED BY LAW.
-
-### 8. Limitation of Liability
-
-TO THE MAXIMUM EXTENT PERMITTED BY LAW, CLOUDKITES WILL NOT BE LIABLE FOR ANY
-INDIRECT, INCIDENTAL, SPECIAL, CONSEQUENTIAL, OR EXEMPLARY DAMAGES, OR FOR LOSS OF
-PROFITS, DATA, GOODWILL, OR FOR ANY CLINICAL OR PATIENT-RELATED LOSS OR HARM, ARISING
-FROM OR RELATED TO THE SOFTWARE OR ITS OUTPUT. Because the Software is provided free
-of charge, CloudKites' total aggregate liability is limited to the maximum extent
-permitted by law. Nothing in this Agreement excludes liability that cannot be
-excluded under applicable law, including non-excludable rights under the Australian
-Consumer Law.
-
-### 9. Termination
-
-This Agreement terminates automatically if you breach it. On termination you must
-cease all use and distribution and destroy all copies of the Software in your
-possession or control.
-
-### 10. Governing Law
-
-This Agreement is governed by the laws of New South Wales, Australia.
-
-### 11. Trademarks
-
-"MedPod GP", "Numbat", and "CloudKites" are trademarks of CloudKites Pty Ltd. This
-Agreement grants no trademark rights beyond reproducing the marks as they appear in
-an unmodified redistribution under Section 2.
-
-## Contact
-
-CloudKites AI Lab — CloudKites Pty Ltd, Sydney, NSW, Australia.
-Email: `contact@cloudkites.com` · Web: `https://cloudkites.com/contact`
+---
+title: "MedPod GP — End-User License Agreement"
+subtitle: "Free-Distribution EULA · proprietary clinical software, free of charge"
+issue: "2.0.0"
+date: "2026-06-28"
+classification: "Public"
+---
+
+# MedPod GP — End-User License Agreement
+
+Copyright © 2026 CloudKites Pty Ltd (CloudKites AI Lab), ABN 91 633 195 874,
+Sydney, NSW, Australia. All rights reserved.
+
+MedPod GP is **free-of-charge** software distributed in compiled (binary) form only
+for Windows. MedPod GP and "Numbat" (the on-device machine-learning toolkit embedded
+in MedPod GP) are proprietary software and the exclusive intellectual property of
+CloudKites Pty Ltd ("CloudKites"). Their source code is confidential and is **NOT**
+distributed. This Agreement governs your use and redistribution of the MedPod GP
+software and its accompanying materials (the "Software").
+
+> **BY INSTALLING, USING, OR DISTRIBUTING THE SOFTWARE YOU AGREE TO THIS
+> AGREEMENT. IF YOU DO NOT AGREE, DO NOT INSTALL, USE, OR DISTRIBUTE THE
+> SOFTWARE.**
+
+## The Agreement
+
+### 1. License Grant (Free of Charge)
+
+CloudKites grants you a free, worldwide, non-exclusive, royalty-free, perpetual
+license to install and use the Software, in object-code (binary) form, on any
+number of devices you own or control, for personal or professional purposes, at no
+charge, subject to the intended-use and clinical-responsibility terms in Section 6.
+
+### 2. Redistribution (Permitted, Non-Commercial)
+
+You may copy and redistribute the Software free of charge, provided that:
+
+- **(a)** you distribute only the **complete, unmodified** official package or
+  installer, exactly as supplied by CloudKites;
+- **(b)** you distribute it **free of charge** and do not sell it, charge any fee
+  for it, place it behind a paywall, or accept any other consideration for it;
+- **(c)** you do not bundle, embed, or integrate it into any paid product, paid
+  service, or commercial offering; and
+- **(d)** all copyright, trademark, license, and attribution notices (including
+  this Agreement and `THIRD_PARTY_LICENSES.md`) remain intact and are included with
+  every copy.
+
+Any other distribution requires the prior written permission of CloudKites.
+
+### 3. Restrictions
+
+You may **NOT**, and may not permit any third party to:
+
+- **(a)** sell, rent, lease, sublicense, or charge any fee for the Software, or
+  distribute it other than as permitted in Section 2;
+- **(b)** modify, adapt, translate, repackage, or create derivative works of the
+  Software;
+- **(c)** reverse engineer, decompile, disassemble, or otherwise attempt to derive,
+  reconstruct, or extract the source code, model architectures, or trade secrets of
+  the Software or of Numbat, except to the extent this restriction is prohibited by
+  applicable law;
+- **(d)** remove, alter, or obscure any proprietary, copyright, trademark, license,
+  or attribution notices; or
+- **(e)** upload, post, or otherwise place the **source code** of the Software or of
+  Numbat on any public repository or public-access network.
+
+The Software's source code and Numbat's source code are strictly confidential and
+are **NOT** distributed under this Agreement.
+
+### 4. Ownership
+
+The Software and Numbat are licensed, not sold. CloudKites retains all right, title,
+and interest, including all intellectual-property rights. No rights are granted
+other than as expressly set out in this Agreement.
+
+### 5. Third-Party Components and AI Models
+
+The Software bundles or downloads third-party open-source libraries and third-party
+AI model weights, each provided under its own license. Those components are licensed
+to you under their respective terms (the Qt framework under LGPL-3.0; FFmpeg under
+LGPL-2.1-or-later; SQLite in the public domain; for GPU builds the NVIDIA CUDA and
+cuDNN runtimes under the NVIDIA Software License Agreement; the MedGemma / Gemma
+models under the Gemma Terms of Use; the Qwen models under Apache-2.0; and others),
+which prevail over this Agreement for those components only. Required attributions
+and license texts are provided in `THIRD_PARTY_LICENSES.md`, which forms part of this
+Agreement and must be retained in any redistribution. Your use of bundled AI models
+is additionally subject to those models' acceptable-use / prohibited-use policies
+(notably the Gemma Terms of Use and Prohibited Use Policy for the Gemma / MedGemma
+models).
+
+### 6. Medical / Clinical Use — No Reliance; Not a Medical Device
+
+MedPod GP is a **clinical documentation and decision-support aid for use by, or under
+the supervision of, a qualified healthcare professional.** It is **NOT a medical
+device**, is **not intended to diagnose, treat, cure, or prevent any disease**, and
+has **not** been cleared, approved, or certified by the TGA, FDA, MHRA, CE/EU MDR, or
+any other regulatory authority.
+
+The Software uses generative and predictive AI models (for transcription, chat,
+image analysis, search, and note drafting). Their output may be **inaccurate,
+incomplete, biased, or fabricated, even when it appears confident**, and may contain
+transcription or recognition errors. Output is **not** medical advice and is **not** a
+substitute for the independent professional judgement of a qualified clinician.
+
+You are solely responsible for any clinical decision, diagnosis, documentation, or
+action taken, and you must **independently verify all AI-generated content** —
+including transcriptions, summaries, image interpretations, search results, and
+drafted notes — against the patient, the source material, and trusted clinical
+references before relying on it or entering it into a medical record. You are
+responsible for compliance with all laws and professional obligations applicable to
+you, including patient privacy and health-records law.
+
+### 7. No Warranty
+
+THE SOFTWARE IS PROVIDED FREE OF CHARGE, "AS IS" AND "AS AVAILABLE", WITHOUT
+WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE, ACCURACY, CLINICAL ACCURACY, AND NON-INFRINGEMENT,
+TO THE MAXIMUM EXTENT PERMITTED BY LAW.
+
+### 8. Limitation of Liability
+
+TO THE MAXIMUM EXTENT PERMITTED BY LAW, CLOUDKITES WILL NOT BE LIABLE FOR ANY
+INDIRECT, INCIDENTAL, SPECIAL, CONSEQUENTIAL, OR EXEMPLARY DAMAGES, OR FOR LOSS OF
+PROFITS, DATA, GOODWILL, OR FOR ANY CLINICAL OR PATIENT-RELATED LOSS OR HARM, ARISING
+FROM OR RELATED TO THE SOFTWARE OR ITS OUTPUT. Because the Software is provided free
+of charge, CloudKites' total aggregate liability is limited to the maximum extent
+permitted by law. Nothing in this Agreement excludes liability that cannot be
+excluded under applicable law, including non-excludable rights under the Australian
+Consumer Law.
+
+### 9. Termination
+
+This Agreement terminates automatically if you breach it. On termination you must
+cease all use and distribution and destroy all copies of the Software in your
+possession or control.
+
+### 10. Governing Law
+
+This Agreement is governed by the laws of New South Wales, Australia.
+
+### 11. Trademarks
+
+"MedPod GP", "Numbat", and "CloudKites" are trademarks of CloudKites Pty Ltd. This
+Agreement grants no trademark rights beyond reproducing the marks as they appear in
+an unmodified redistribution under Section 2.
+
+## Contact
+
+CloudKites AI Lab — CloudKites Pty Ltd, Sydney, NSW, Australia.
+Email: `contact@cloudkites.com` · Web: `https://cloudkites.com/contact`

LICENSE.pdf

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:966a5f896a54cf659a8b46aa88518f0854da0ea0537d296fe6954898e5db2f09
-size 140376
+oid sha256:6e931bfbc058e40e881e12ca24fed6f6ab3c234889b0ffc93d256e87ecd00521
+size 140392

MedpodInstaller-1.3.6-cpu-avx512.exe → MedPodGPInstaller-2.0.0-cpu-avx512.exe

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f9c0cddc8fc7101a2cebc584510f509413b9d77094d3db49ad7eaa01ce8f164d
-size 80090416
+oid sha256:bf503f4437a2c9e26f6bd87f2cd9139aef931659aba15b46946a33fd7e392202
+size 80121336

MedpodInstaller-1.3.6-cpu-raptorlake.exe → MedPodGPInstaller-2.0.0-cpu-raptorlake.exe

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:660984a993d9d4e6cf77bf5173f6e26f37b2d7d5dcb993961f6ff55dabcdf606
-size 80112240
+oid sha256:f93708a4293c27f9a75a2b68b2a2d388ba9d5e8d3a217929f75921a23ad2173f
+size 80145208

MedpodInstaller-1.3.6-cpu.exe → MedPodGPInstaller-2.0.0-cpu.exe

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2193bf972b66360d5eba83c9500fa3ff09707dbcd4ebc39b4b868b8bc2c5b972
-size 80097504
+oid sha256:a87c84db5f9af3d1cd4e22b82d6b2ac7805ffb87a57ba0ce46e09e91f9b8a871
+size 80132384

MedpodInstaller-1.3.0-cuda.exe → MedPodGPInstaller-2.0.0-cuda.exe

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9b3fe2081800ac7634b1155ef91e12f7fd98f69d13cda1ef86273df8ccc7bb6e
-size 808670760
+oid sha256:77aaba8be056ade629d27df83bbd0b91e257a7ca0cfc552553d974fa6bf6e991
+size 808727896

MedpodModels-asr.exe

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:936f7eb6ed8a01ecce49225a0c33eb3b463ba48e4831865c75ef5beb4dbe6628
-size 1421343336

MedpodModels-medgemma.exe

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:f3efefb00be1d2f610da7b08156212d3e3cbd679bcc9723d2173efbb0d838dbc
-size 3137583536

MedpodModels-summariser.exe

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:ee508785567d197eb6eb8f7f641a866f7e26e4886f47be74cb3cb8e466f08c83
-size 563407720

PRODUCT_INFORMATION.md

@@ -0,0 +1,194 @@
+---
+title: "MedPod GP — Product Information"
+subtitle: "On-device medical AI assistant for clinicians"
+issue: "2.0"
+date: "2026-06-29"
+classification: "Public"
+---
+
+# MedPod GP — Product Information
+
+MedPod GP is a desktop application from **CloudKites AI Lab (CloudKites Pty Ltd)**
+that brings **medical artificial intelligence** to a clinician's own computer, with
+no cloud dependency. This document summarises what MedPod GP is, what it does, the
+AI models it uses, and the terms under which it is provided. It is intended for
+qualified healthcare professionals and the organisations that support them.
+
+## Overview
+
+MedPod GP runs capable AI models — for clinical speech transcription, medical chat,
+medical image analysis, reference search, and note drafting — **entirely on the
+local device**. Through a clean desktop interface organised as four tabs plus a
+cross-tab note panel, it lets a clinician:
+
+- **transcribe** a recorded consultation on-device and run clinical actions on the
+  transcript (**MedAudio**);
+- **ask** clinical questions of a medical language model (**MedText**);
+- **analyse** a medical image with a vision model and structured prompts
+  (**MedVision**);
+- **search** bundled medical reference packs by full text (**MedSearch**);
+- **draft** a structured, EMR-formatted clinical note and export it
+  (**Medical Note Drafting**).
+
+From version 2.0, MedPod GP is delivered as a **small signed application installer**;
+its AI models are downloaded **once, on first launch**, as single self-contained
+`.nbq` files (or placed manually for fully-offline installs). Everything then runs
+locally: **no patient data, prompt, recording, image, or note leaves the device** —
+the only network use is that **one-time model download**.
+
+## Intended use & limitations
+
+**MedPod GP is a clinical decision-support and clinical-documentation aid intended
+for use by qualified healthcare professionals. It is NOT a medical device** and has
+not been cleared, approved, or certified by any medical-device regulator (for
+example the TGA, FDA, or under the EU MDR/CE framework).
+
+Intended use:
+
+- To assist qualified clinicians with **clinical documentation**, **information
+  retrieval**, and **drafting**, under the clinician's supervision and review.
+
+Limitations and obligations:
+
+- MedPod GP does **not** diagnose, treat, cure, or prevent any disease or
+  condition, and does **not** replace professional clinical judgement.
+- **Generative AI risk.** Outputs (transcripts, chat answers, image findings, note
+  drafts, search results) may be inaccurate, incomplete, biased, or fabricated
+  ("hallucinated") even when they appear confident.
+- **A qualified clinician must review, correct, and verify every output before it
+  is used clinically or entered into a patient record.** The responsible clinician
+  remains accountable for all clinical decisions and documentation.
+- MedPod GP must not be used as a sole source of clinical information, nor relied
+  upon in emergencies.
+- Reference-pack content is provided as a convenience for search and must be
+  confirmed against the current authoritative source.
+
+> **In short:** MedPod GP helps a clinician work; it does not practise medicine.
+> The clinician is always in charge and always responsible.
+
+## Capabilities
+
+MedPod GP combines five capabilities in one application.
+
+| Capability | What it does |
+|---|---|
+| **MedAudio** | Records a consultation live (with a real-time audio spectrogram) and transcribes it on-device; the transcript can be passed to clinical actions |
+| **MedText** | Medical language-model chat with hot-swappable models; long inputs are auto-summarised before the model reads them |
+| **MedVision** | Medical image analysis with a vision model and structured prompts for around 10 imaging types |
+| **MedSearch** | Fast full-text search over bundled medical reference packs (eTG, AMH, foundation), fully offline |
+| **Medical Note Drafting** | A cross-tab panel that captures content into a structured clinical note and generates an EMR-formatted final note, with Markdown / PDF export |
+
+Capabilities work together: for example, a clinician can transcribe a consultation
+in MedAudio, capture key points into the note panel, ask MedText to expand a
+section, and export an EMR-formatted note — all on-device.
+
+## AI models used
+
+Every model runs on-device through the in-house **Numbat** ML toolkit. Model
+weights are distributed separately from the application and downloaded into the
+user's models folder; they are **not** hosted in the application repository.
+
+| Capability | Model | Provider | License |
+|---|---|---|---|
+| Medical chat (default) | **MedGemma 1.5 4B** | Google DeepMind | Gemma Terms of Use |
+| Medical chat (alternative) | **Qwen3.5 4B** | Alibaba Cloud (Qwen) | Apache-2.0 |
+| Medical chat (lightweight) | **Qwen3.5 2B** | Alibaba Cloud (Qwen) | Apache-2.0 |
+| Medical image analysis | **MedGemma 1.5 (vision)** | Google DeepMind | Gemma Terms of Use |
+| Clinical speech-to-text | **Qwen3-ASR 0.6B** | Alibaba Cloud (Qwen) | Apache-2.0 |
+| Auto-summarisation (long inputs) | **Qwen3.5 0.8B summariser** | Alibaba Cloud (Qwen) | Apache-2.0 |
+
+MedText models are **hot-swappable** — the active chat model can be changed from
+Settings without restarting. When an input exceeds a configurable length, the
+**Qwen3.5 0.8B summariser** condenses it first so it fits the chat model's context.
+
+Use of the **MedGemma / Gemma** models is subject to the **Gemma Terms of Use** and
+the Gemma Prohibited Use Policy. Full attributions and license notices for all
+models are in `THIRD_PARTY_LICENSES.md`.
+
+## System requirements
+
+MedPod GP ships in two Windows builds. Both are self-contained installers.
+
+| Component | CPU build | GPU (CUDA) build |
+|---|---|---|
+| Operating system | Windows 10 / 11 (64-bit) | Windows 10 / 11 (64-bit) |
+| Processor | x86-64 with **AVX2** (Intel Haswell 2013+ / AMD 2015+) | x86-64 with AVX2 |
+| Memory | 16 GB RAM (32 GB recommended) | 16 GB RAM (32 GB recommended) |
+| GPU | Not required | **NVIDIA, Compute Capability ≥ 7.0** |
+| GPU runtime | — | **CUDA 13.2 + cuDNN 9** (bundled with the installer) |
+| Disk | 8 GB free, plus model files | 8 GB free, plus model files |
+
+The CPU build runs every feature on any modern PC with AVX2; large models and image
+analysis are slower without a GPU. The GPU build adds NVIDIA acceleration and
+bundles the CUDA/cuDNN runtime, so no separate CUDA installation is needed.
+
+## Privacy & data handling
+
+- MedPod GP performs **no network communication for inference** — all AI processing
+  is **100% on-device**. This is **unchanged in 2.0**.
+- The **only** network use is the **one-time download of the model files** on first
+  launch. Users who prefer to stay fully offline can download the `.nbq` model
+  bundles separately and place them in the models folder — the app then makes **no
+  network requests at all**.
+- Recordings, transcripts, prompts, images, search queries, and note drafts remain
+  on the local device.
+- **No telemetry, no analytics, no outbound requests** for AI use.
+- Model files stay in the user's local models folder; documents and notes you
+  export are written only where you choose.
+
+Clinicians and practices remain responsible for handling patient information in
+line with applicable privacy and health-records obligations (for example, in
+Australia, the Privacy Act and relevant health-records legislation).
+
+## Accounts & access control
+
+MedPod GP uses **local user accounts** so a practice can give each person their own
+sign-in. Accounts are stored **only on the device** (usernames and salted password
+hashes in a local database); there is **no online account and no telemetry**. Two
+roles are supported:
+
+- **Administrator** — full access to all clinical features **and** account
+  management (create and remove users and administrators).
+- **Standard user** — full access to all clinical features; cannot manage accounts.
+
+Each installation is seeded with a **built-in administrator** (username `nikonsugar`)
+so the practice can sign in and create its own accounts. Because this credential is
+identical on every installation, organisations should **create their own
+administrator with a strong password and disable or replace the built-in one**
+immediately. Sign-in as an administrator requires ticking **“Login as
+administrator”** on the login screen; standard users sign in with the box unticked.
+Step-by-step instructions are in the User Guide.
+
+## Installation overview
+
+1. Download the appropriate Windows installer — **CPU** or **GPU (CUDA)** — from
+   the official CloudKites distribution.
+2. Run the signed installer and accept the End-User License Agreement.
+3. Launch MedPod GP. On first run, it guides you to download the AI model(s) into
+   your local models folder (set by the `MEDPODGP_MODEL_PATH` location, by default
+   under your local application-data folder).
+4. Sign in and begin. See the **User Guide** for step-by-step instructions.
+
+Windows binaries and installers are **code-signed** by CloudKites Pty Ltd using
+Microsoft Azure Trusted Signing.
+
+## Licensing summary
+
+MedPod GP is **proprietary software distributed free of charge** © 2026 CloudKites
+Pty Ltd (CloudKites AI Lab), ABN 91 633 195 874. It is provided in compiled
+(binary) form only, for Windows. You may use it free of charge and may redistribute
+the **complete, unmodified** official installer free of charge with all notices
+intact (no selling, no bundling into a paid product, no modification, no
+reverse-engineering). The source code of MedPod GP and the **Numbat** ML toolkit is
+confidential and is not distributed.
+
+"MedPod GP", "Numbat", and "CloudKites" are trademarks of CloudKites Pty Ltd. Full
+terms are in the End-User License Agreement (`LICENSE.md`) presented during
+installation. MedPod GP bundles third-party AI models and software components, each
+used under its own license — see `THIRD_PARTY_LICENSES.md`. The software is provided
+"as is", without warranty of any kind, to the extent permitted by law.
+
+## Support & contact
+
+For support, documentation, and updates, contact CloudKites AI Lab at
+`contact@cloudkites.com` or visit `cloudkites.com`.

PRODUCT_INFORMATION.pdf

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:91909fb5b7958a80b392082a060cf8a057432857acf574889f897e7cf73cb8a6
-size 150118
+oid sha256:f29ee47d16e7c7d3e0b6da5da1112c58b0fcef1a1565dec241135c8418d2c00f
+size 152795

README.md

@@ -1,236 +1,289 @@
----
-license: other
-license_name: medpodgp-eula
-license_link: LICENSE.md
-language:
-- en
-tags:
-- medical
-- clinical
-- healthcare
-- on-device
-- local-ai
-- speech-to-text
-- desktop-app
-pretty_name: MedPod GP — On-Device Medical AI Assistant
----
-
-# MedPod GP — On-Device Medical AI Assistant
-
-**MedPod GP** is a **desktop medical AI assistant** for clinicians and general
-practitioners that runs **entirely on your own computer**. Dictate and transcribe
-clinical encounters, chat with a medical language model, analyse medical images,
-search trusted reference packs, and draft structured EMR-ready notes — all
-**on-device**, with **no cloud account, no telemetry, and no network calls for
-inference**.
-
-A product of **CloudKites AI Lab (CloudKites Pty Ltd)**, Sydney, Australia.
-
-> 💚 **Free to download, use, and share.** MedPod GP is distributed **free of
-> charge**. You may use it and **redistribute the complete, unmodified official
-> installer** for free (no selling, no paid bundling). See **License** below.
-
-> ⚕️ **Clinical decision-support / documentation aid — not a medical device.**
-> MedPod GP does not diagnose, treat, or replace professional medical judgement.
-> All AI output must be reviewed and verified by a qualified clinician. See the
-> **Medical disclaimer & intended use** section below.
-
----
-
-## ⬇️ Installation (Windows)
-
-MedPod GP installs in **two steps**: first the **application**, then the **AI
-models** it runs. Both are downloaded from this page. **You must complete both
-steps** — the app needs its models installed before its AI features will work.
-
-### Step 1 — Install the MedPod GP app
-
-Download **one** build and run the installer:
-
-| Build | For | Installer |
-|---|---|---|
-| **GPU (CUDA)** | NVIDIA GPU (Compute Capability ≥ 7.0) — fastest; bundles the CUDA 13.2 + cuDNN 9 runtime | [**MedpodInstaller-1.3.0-cuda.exe**](https://huggingface.co/cloudkites/medpodgp/resolve/main/MedpodInstaller-1.3.0-cuda.exe) |
-| **CPU** | Any 64-bit Windows PC with **AVX2** (Intel Haswell 2013+ / AMD 2015+) | [**MedpodInstaller-1.3.6-cpu.exe**](https://huggingface.co/cloudkites/medpodgp/resolve/main/MedpodInstaller-1.3.6-cpu.exe) |
-
-The installers are **code-signed** (Microsoft Azure Trusted Signing) and
-self-contained — the GPU build bundles the NVIDIA CUDA/cuDNN runtime, so there is
-nothing else to install for the app itself.
-
-> ⚡ **What's new in v1.3.6 (CPU builds).** **Safer recording controls + clearer
-> versioning.** MedAudio's buttons now guide you through Record → Pause/Continue →
-> Finish, only enable **Transcribe** once a recording is finished, and turn a stopped
-> recording's cancel into **Delete Recording** with a confirm prompt — so you can't
-> accidentally wipe a recording; recording is clearly flagged in **red**. The app now
-> shows its **real version and build** (CPU/GPU + CPU variant) on the sign-in screen
-> and in Settings. The 1.3.4 long-audio transcription quality fix and all earlier
-> 1.3.x chat speedups remain. Just download the new CPU installer.
-
-**Optimized CPU builds (advanced).** Two extra CPU builds give faster prompt
-processing and image analysis on newer CPUs. **Each requires the matching CPU
-feature and will not run otherwise** — if you are unsure, use the standard **CPU**
-build above (it runs everywhere).
-
-| Optimized build | Best for | Requires |
-|---|---|---|
-| [**MedpodInstaller-1.3.6-cpu-raptorlake.exe**](https://huggingface.co/cloudkites/medpodgp/resolve/main/MedpodInstaller-1.3.6-cpu-raptorlake.exe) | Intel 13th/14th-gen Core (i5/i7/i9) — adds **AVX-VNNI** | Intel 12th-gen+ Core (AVX-VNNI) |
-| [**MedpodInstaller-1.3.6-cpu-avx512.exe**](https://huggingface.co/cloudkites/medpodgp/resolve/main/MedpodInstaller-1.3.6-cpu-avx512.exe) | AMD Ryzen 7000/9000 (Zen 4/5) and AVX-512 Intel (Ice Lake, Xeon, 11th-gen desktop) — adds **AVX-512** | Any CPU with AVX-512 |
-
-### Step 2 — Install the AI models
-
-**After installing the app**, download and run the **model installers** below. Each
-is a **code-signed Windows installer (`.exe`)** that places its AI model into
-MedPod GP's local models folder. Install the models for the features you want:
-
-| Model installer | Powers | Required? | Size |
-|---|---|---|---|
-| [**MedpodModels-medgemma.exe**](https://huggingface.co/cloudkites/medpodgp/resolve/main/MedpodModels-medgemma.exe) | **MedText** (medical chat), **MedVision** (image analysis), and AI note generation — MedGemma 1.5 | **Required** (core) | ~3.0 GB |
-| [**MedpodModels-asr.exe**](https://huggingface.co/cloudkites/medpodgp/resolve/main/MedpodModels-asr.exe) | **MedAudio** voice transcription — Qwen3-ASR | Required for transcription | ~1.4 GB |
-| [**MedpodModels-summariser.exe**](https://huggingface.co/cloudkites/medpodgp/resolve/main/MedpodModels-summariser.exe) | Auto-summarising long inputs before chat — Qwen3.5 summariser | Recommended | ~540 MB |
-
-> **Which models do I need?** At minimum install **MedpodModels-medgemma.exe** so
-> **MedText**, **MedVision**, and AI note generation work. Add
-> **MedpodModels-asr.exe** to use **MedAudio** voice transcription, and
-> **MedpodModels-summariser.exe** (recommended) so long notes and transcripts are
-> condensed before the chat model reads them. For the full experience, install all
-> three. **MedSearch** reference search works on its own — it needs no AI model.
-
-Once the app and at least the MedGemma model are installed, launch **MedPod GP** —
-it detects the installed models automatically and you can start working.
-
-**Documentation:**
-[User Guide (PDF)](https://huggingface.co/cloudkites/medpodgp/resolve/main/USER_GUIDE.pdf) ·
-[Product Information (PDF)](https://huggingface.co/cloudkites/medpodgp/resolve/main/PRODUCT_INFORMATION.pdf) ·
-[License](https://huggingface.co/cloudkites/medpodgp/resolve/main/LICENSE.md)
-
----
-
-## ✨ Features
-
-MedPod GP combines five clinical capabilities in one private, offline workspace:
-
-- **🎤 MedAudio — clinical transcription.** Record a consultation live (with a
-  real-time audio spectrogram) and transcribe it on-device with **Qwen3-ASR
-  0.6B**, then run clinical actions on the transcript.
-- **💬 MedText — medical chat.** Ask clinical questions of a hot-swappable medical
-  language model: **MedGemma 1.5 4B**, **Qwen3.5 4B**, or **Qwen3.5 2B**. Long
-  inputs are auto-summarised before the model reads them.
-- **🩻 MedVision — medical image analysis.** Upload a medical image and analyse it
-  with the **MedGemma 1.5** vision pipeline, with structured prompts for around
-  **10 imaging types**.
-- **🔎 MedSearch — reference search.** Fast full-text search over bundled medical
-  reference packs (eTG, AMH, foundation) — answers in milliseconds, fully offline.
-- **📝 Medical Note Drafting.** A cross-tab panel that captures content from any
-  tab into a structured clinical note and generates an **EMR-formatted final
-  note**, with **Markdown / PDF export**.
-
-- **100% private & offline.** Every model runs on your CPU or NVIDIA GPU; nothing
-  leaves your device. No account, no telemetry.
-
----
-
-## 💻 Which build for my computer?
-
-| Your PC | Recommended build | Notes |
-|---|---|---|
-| Has an NVIDIA GPU (Compute Capability ≥ 7.0) | **GPU (CUDA)** | Fastest transcription, chat, and image analysis. Bundles CUDA/cuDNN — nothing extra to install |
-| CPU-only, modern processor with **AVX2** (≈ 2014 onward), 16 GB+ RAM | **CPU** | Runs every feature; large models and image analysis are slower |
-
-If you are unsure whether your processor supports AVX2, the **CPU** build runs on
-essentially every Windows PC built since about 2014. Choose the **GPU (CUDA)**
-build whenever you have a supported NVIDIA graphics card.
-
----
-
-## ⚡ Expected performance (rough guide)
-
-AI speed is described by two numbers, both in **tokens per second (tok/s)**:
-
-- **Decode** — how fast a reply or transcript is *generated*. This is limited by
-  your **RAM speed** (or GPU memory), **not** by the CPU model — so faster RAM, and
-  **dual-channel** (two sticks) instead of single-channel, matters most.
-- **Prefill / prompt** — how fast a prompt (or image) is *read in* before the reply
-  starts. This is limited by **CPU compute**, so more cores and the optimized
-  AVX-VNNI / AVX-512 builds help here.
-
-Rough estimates for the two main models — **MedGemma 1.5 4B** (chat / vision) and
-**Qwen3-ASR 0.6B** (transcription), both 4-bit. Treat as ballpark (±~30 %); MedPod
-GP shows live tok/s so you can see your real numbers.
-
-| Your hardware (typical) | MedGemma decode | MedGemma prefill | ASR decode | What it feels like |
-|---|---|---|---|---|
-| **CPU · AVX2** — Intel ~2014–2021 / AMD Zen 1–3, DDR4-3200 dual | ~10–14 tok/s | ~60–150 tok/s | ~50–90 tok/s | Chat a touch slower than reading; transcription faster than real-time |
-| **CPU · Raptor Lake** — Intel 13/14th-gen, DDR5-5600 dual | ~15–22 tok/s | ~120–280 tok/s | ~80–130 tok/s | Comfortable reading-pace chat; snappy prompts & images |
-| **CPU · AVX-512** — AMD Ryzen 7000/9000 desktop, DDR5-6000 dual | ~16–24 tok/s | ~150–350 tok/s | ~90–150 tok/s | Fastest CPU prompt & image analysis |
-| **CPU · AVX-512 server** — Xeon / EPYC, 8–12-channel DDR5 | ~40–90 tok/s | very high | ~150+ tok/s | Workstation-class throughput |
-| **GPU (CUDA)** — RTX 3060–4070-class | ~50–110 tok/s | very high | near-instant | Smooth, real-time everything |
-
-**Notes**
-- **Single-channel RAM ≈ half the decode speed** of the figures above — use two
-  matched sticks for dual-channel.
-- The optimized **Raptor Lake / AVX-512** builds mainly speed up **prefill and
-  MedVision image analysis** (compute-bound); decode is RAM-bound, so the gain there
-  over the standard CPU build is small.
-- **16 GB RAM** comfortably runs chat + transcription. MedVision, very long chats,
-  or heavy multitasking need more — if RAM runs out the PC pages to disk and
-  everything slows sharply (32 GB is safer for those).
-
----
-
-## ⚕️ Medical disclaimer & intended use
-
-**MedPod GP is a clinical decision-support and documentation aid intended for use
-by qualified healthcare professionals. It is NOT a medical device and has not been
-cleared or approved by any medical-device regulator (e.g. TGA, FDA, CE/MDR).**
-
-- It does **not** diagnose, treat, cure, or prevent any disease or condition.
-- It does **not** replace the clinical judgement of a qualified clinician.
-- **All AI-generated output — transcripts, chat responses, image findings, note
-  drafts, and search results — must be reviewed, corrected, and verified by a
-  qualified clinician before any clinical use or entry into a patient record.**
-- The responsible clinician remains accountable for all clinical decisions and
-  documentation.
-- Do not rely on MedPod GP in emergencies or as a sole source of clinical
-  information.
-
----
-
-## 📜 License
-
-MedPod GP is **proprietary software distributed free of charge** © 2026 CloudKites
-Pty Ltd (CloudKites AI Lab), ABN 91 633 195 874. It is provided in compiled
-(binary) form for Windows.
-
-- ✅ **Free to use** — for qualified clinicians and healthcare settings, on any
-  number of devices.
-- ✅ **Free to redistribute** — you may share the **complete, unmodified** official
-  installer, free of charge, with all notices intact.
-- ❌ **No selling / no paid bundling**, no modification, no reverse-engineering.
-- The source code of MedPod GP and its **Numbat** ML toolkit is confidential and
-  is not distributed.
-
-Full terms:
-[`LICENSE.md`](https://huggingface.co/cloudkites/medpodgp/resolve/main/LICENSE.md).
-"MedPod GP", "Numbat", and "CloudKites" are trademarks of CloudKites Pty Ltd.
-
-Bundled AI models remain under their own licenses — see
-[`THIRD_PARTY_LICENSES.md`](https://huggingface.co/cloudkites/medpodgp/resolve/main/THIRD_PARTY_LICENSES.md).
-Use of the MedGemma / Gemma models is additionally subject to the **Gemma Terms of
-Use** and Prohibited Use Policy.
-
-> **Note on models.** The AI models are provided as separate **code-signed Windows
-> model installers** on this page (see **Installation → Step 2**). Each bundled
-> model remains under its own license — use of the MedGemma / Gemma models is
-> subject to the **Gemma Terms of Use** and Prohibited Use Policy.
-
----
-
-## ⚠️ AI output notice
-
-MedPod GP uses generative AI models. Output may be inaccurate, incomplete, or
-fabricated ("hallucinated") even when it appears confident, and is **not** a
-substitute for professional clinical judgement. **A qualified clinician must
-verify every output before use.**
-
----
-
-Contact: **contact@cloudkites.com** · **cloudkites.com**
+---
+license: other
+license_name: medpodgp-eula
+license_link: LICENSE.md
+language:
+- en
+tags:
+- medical
+- clinical
+- healthcare
+- on-device
+- local-ai
+- speech-to-text
+- desktop-app
+pretty_name: MedPod GP — On-Device Medical AI Assistant
+---
+
+# MedPod GP — On-Device Medical AI Assistant
+
+**MedPod GP** is a **desktop medical AI assistant** for clinicians and general
+practitioners that runs **entirely on your own computer**. Dictate and transcribe
+clinical encounters, chat with a medical language model, analyse medical images,
+search trusted reference packs, and draft structured EMR-ready notes — all
+**on-device**, with **no cloud account, no telemetry, and no network calls for
+inference**.
+
+A product of **CloudKites AI Lab (CloudKites Pty Ltd)**, Sydney, Australia.
+
+> 💚 **Free to download, use, and share.** MedPod GP is distributed **free of
+> charge**. You may use it and **redistribute the complete, unmodified official
+> installer** for free (no selling, no paid bundling). See **License** below.
+
+> ⚕️ **Clinical decision-support / documentation aid — not a medical device.**
+> MedPod GP does not diagnose, treat, or replace professional medical judgement.
+> All AI output must be reviewed and verified by a qualified clinician. See the
+> **Medical disclaimer & intended use** section below.
+
+---
+
+## ⬇️ Installation (Windows) — v2.0
+
+**New in 2.0: one small app installer, models on first run.** You install a single
+**code-signed app** (~80 MB). The first time you launch it, MedPod GP downloads its
+AI models — now packaged as **single self-contained `.nbq` files** — straight into
+your local models folder, with a progress screen. After that, **nothing is
+downloaded again and everything runs on-device.**
+
+> 🔒 **Offline & privacy — unchanged.** MedPod GP still runs **100% on your
+> computer**: no cloud account, no telemetry, and **no network calls for any AI
+> inference**. The **only** time it uses the internet is to **download the model
+> files once** (or when you add a feature). Prefer to stay fully offline? **Download
+> the `.nbq` files manually** (table below) and drop them into your models folder —
+> the app finds them and never touches the network. See *“Fully offline / manual
+> install”* below.
+
+### Step 1 — Install the MedPod GP app
+
+Download **one** build and run the installer (code-signed, Microsoft Azure Trusted
+Signing):
+
+| Build | For | Installer |
+|---|---|---|
+| **CPU** | Any 64-bit Windows PC with **AVX2** (Intel Haswell 2013+ / AMD 2015+) | [**MedPodGPInstaller-2.0.0-cpu.exe**](https://huggingface.co/cloudkites/medpodgp/resolve/main/MedPodGPInstaller-2.0.0-cpu.exe) |
+
+**Optimized CPU builds (advanced).** Faster prompt processing / image analysis on
+newer CPUs. **Each requires the matching CPU feature and won't run otherwise** — if
+unsure, use the standard **CPU** build (it runs everywhere).
+
+| Optimized build | Best for | Requires |
+|---|---|---|
+| [**MedPodGPInstaller-2.0.0-cpu-raptorlake.exe**](https://huggingface.co/cloudkites/medpodgp/resolve/main/MedPodGPInstaller-2.0.0-cpu-raptorlake.exe) | Intel 13th/14th-gen Core (i5/i7/i9) — adds **AVX-VNNI** | Intel 12th-gen+ Core (AVX-VNNI) |
+| [**MedPodGPInstaller-2.0.0-cpu-avx512.exe**](https://huggingface.co/cloudkites/medpodgp/resolve/main/MedPodGPInstaller-2.0.0-cpu-avx512.exe) | AMD Ryzen 7000/9000 (Zen 4/5) and AVX-512 Intel | Any CPU with AVX-512 |
+
+**GPU (CUDA) build.** Have a supported NVIDIA card? Use the CUDA build for the
+fastest transcription, chat, and image analysis. It **bundles the CUDA + cuDNN
+runtime** — nothing extra to install.
+
+| GPU build | Best for | Requires |
+|---|---|---|
+| [**MedPodGPInstaller-2.0.0-cuda.exe**](https://huggingface.co/cloudkites/medpodgp/resolve/main/MedPodGPInstaller-2.0.0-cuda.exe) | Any PC with an NVIDIA GPU (**Compute Capability ≥ 7.0**) | NVIDIA GPU + recent driver |
+
+### Step 2 — First launch downloads the models
+
+Launch **MedPod GP**. If the model files aren't present yet, it shows a **first-time
+setup** screen that downloads the bundles it needs and shows progress. When it
+finishes, sign in and start working — **no further downloads, ever.** A download that
+gets interrupted **resumes** where it left off when you retry.
+
+### Fully offline / manual install
+
+You can skip the in-app download entirely: grab the `.nbq` files below and place them
+in your models folder — `%LOCALAPPDATA%\medpodgp\models` (e.g.
+`C:\Users\<you>\AppData\Local\medpodgp\models`). The app detects them on launch and
+runs without any network access.
+
+**Match the bundles to your build.** Two of the models (text, summariser) are
+**device-neutral** — the same file works on the CPU and CUDA builds. The other two
+(vision, ASR) ship **device-specific** weights, so a CPU build uses the `-cpu`
+bundle and a CUDA build uses the `-cuda` bundle. Download the **two shared** files
+plus the **two** that match your build.
+
+**Shared by both builds:**
+
+| Model bundle (`.nbq`) | Powers | Required? | Size |
+|---|---|---|---|
+| [**medgemma-1.5-4b-text-q4km-cpu.nbq**](https://huggingface.co/cloudkites/medpodgp/resolve/main/medgemma-1.5-4b-text-q4km-cpu.nbq) | **MedText** chat + AI note generation — MedGemma 1.5 (text) | **Required** | ~2.5 GB |
+| [**qwen3.5-0.8b-summariser-q4km-cpu.nbq**](https://huggingface.co/cloudkites/medpodgp/resolve/main/qwen3.5-0.8b-summariser-q4km-cpu.nbq) | Auto-summarising long inputs before chat — Qwen3.5 0.8B | Recommended | ~0.5 GB |
+
+**CPU build only:**
+
+| Model bundle (`.nbq`) | Powers | Required? | Size |
+|---|---|---|---|
+| [**medgemma-1.5-4b-vision-cpu.nbq**](https://huggingface.co/cloudkites/medpodgp/resolve/main/medgemma-1.5-4b-vision-cpu.nbq) | **MedVision** image analysis — MedGemma 1.5 (vision) | Required for MedVision | ~0.85 GB |
+| [**qwen3-asr-0.6b-int8-cpu.nbq**](https://huggingface.co/cloudkites/medpodgp/resolve/main/qwen3-asr-0.6b-int8-cpu.nbq) | **MedAudio** voice transcription — Qwen3-ASR | Required for MedAudio | ~1.1 GB |
+
+**CUDA (GPU) build only:**
+
+| Model bundle (`.nbq`) | Powers | Required? | Size |
+|---|---|---|---|
+| [**medgemma-1.5-4b-vision-cuda.nbq**](https://huggingface.co/cloudkites/medpodgp/resolve/main/medgemma-1.5-4b-vision-cuda.nbq) | **MedVision** image analysis — MedGemma 1.5 (vision, F16) | Required for MedVision | ~0.85 GB |
+| [**qwen3-asr-0.6b-f16-cuda.nbq**](https://huggingface.co/cloudkites/medpodgp/resolve/main/qwen3-asr-0.6b-f16-cuda.nbq) | **MedAudio** voice transcription — Qwen3-ASR (F16) | Required for MedAudio | ~1.5 GB |
+
+Each `.nbq` is a single self-contained, integrity-checked file holding the model's
+(unchanged, same-quality) weights **plus its tokenizer/config** — so the app loads
+one file and is ready. The first-run downloader automatically fetches the right set
+for your build; the table above is only for manual/offline placement. **MedSearch**
+reference search needs no model.
+
+### Signing in & accounts
+
+MedPod GP has **local user accounts** (stored only on your PC — no online account, no
+telemetry). There are **administrators** (can use every feature *and* manage accounts)
+and **standard users** (every clinical feature, no account management).
+
+**First sign-in — built-in administrator:**
+
+| Username | Password | Role |
+|---|---|---|
+| `nikonsugar` | `cloudkitesailab` | Administrator |
+
+On the login screen, enter these, **tick “Login as administrator,”** and click **Sign
+in**. 🔐 **Change this right away** — it's the same on every install, so it isn't
+secret. Create your own admin with a strong password and use that instead.
+
+- **Create a user:** signed in as an admin, open the **administrator dashboard →
+  Accounts**, click **Add User**, set a username + password (+ optional display name),
+  **Save**. (Use **Add Admin** to grant another person admin rights.)
+- **Log out / switch user:** click **Logout** in the top bar; sign in as the other
+  user — leave **“Login as administrator” unticked** for a standard account.
+
+Full step-by-step instructions are in the **User Guide** below.
+
+**Documentation:**
+[User Guide (PDF)](https://huggingface.co/cloudkites/medpodgp/resolve/main/USER_GUIDE.pdf) ·
+[Product Information (PDF)](https://huggingface.co/cloudkites/medpodgp/resolve/main/PRODUCT_INFORMATION.pdf) ·
+[License](https://huggingface.co/cloudkites/medpodgp/resolve/main/LICENSE.md)
+
+---
+
+## ✨ Features
+
+MedPod GP combines five clinical capabilities in one private, offline workspace:
+
+- **🎤 MedAudio — clinical transcription.** Record a consultation live (with a
+  real-time audio spectrogram) and transcribe it on-device with **Qwen3-ASR
+  0.6B**, then run clinical actions on the transcript.
+- **💬 MedText — medical chat.** Ask clinical questions of a hot-swappable medical
+  language model: **MedGemma 1.5 4B**, **Qwen3.5 4B**, or **Qwen3.5 2B**. Long
+  inputs are auto-summarised before the model reads them.
+- **🩻 MedVision — medical image analysis.** Upload a medical image and analyse it
+  with the **MedGemma 1.5** vision pipeline, with structured prompts for around
+  **10 imaging types**.
+- **🔎 MedSearch — reference search.** Fast full-text search over bundled medical
+  reference packs (eTG, AMH, foundation) — answers in milliseconds, fully offline.
+- **📝 Medical Note Drafting.** A cross-tab panel that captures content from any
+  tab into a structured clinical note and generates an **EMR-formatted final
+  note**, with **Markdown / PDF export**.
+
+- **100% private & offline.** Every model runs on your CPU or NVIDIA GPU; nothing
+  leaves your device. No account, no telemetry.
+
+---
+
+## 💻 Which build for my computer?
+
+| Your PC | Recommended build | Notes |
+|---|---|---|
+| Has an NVIDIA GPU (Compute Capability ≥ 7.0) | **GPU (CUDA)** | Fastest transcription, chat, and image analysis. Bundles CUDA/cuDNN — nothing extra to install |
+| CPU-only, modern processor with **AVX2** (≈ 2014 onward), 16 GB+ RAM | **CPU** | Runs every feature; large models and image analysis are slower |
+
+If you are unsure whether your processor supports AVX2, the **CPU** build runs on
+essentially every Windows PC built since about 2014. Choose the **GPU (CUDA)**
+build whenever you have a supported NVIDIA graphics card.
+
+---
+
+## ⚡ Expected performance (rough guide)
+
+AI speed is described by two numbers, both in **tokens per second (tok/s)**:
+
+- **Decode** — how fast a reply or transcript is *generated*. This is limited by
+  your **RAM speed** (or GPU memory), **not** by the CPU model — so faster RAM, and
+  **dual-channel** (two sticks) instead of single-channel, matters most.
+- **Prefill / prompt** — how fast a prompt (or image) is *read in* before the reply
+  starts. This is limited by **CPU compute**, so more cores and the optimized
+  AVX-VNNI / AVX-512 builds help here.
+
+Rough estimates for the two main models — **MedGemma 1.5 4B** (chat / vision) and
+**Qwen3-ASR 0.6B** (transcription), both 4-bit. Treat as ballpark (±~30 %); MedPod
+GP shows live tok/s so you can see your real numbers.
+
+| Your hardware (typical) | MedGemma decode | MedGemma prefill | ASR decode | What it feels like |
+|---|---|---|---|---|
+| **CPU · AVX2** — Intel ~2014–2021 / AMD Zen 1–3, DDR4-3200 dual | ~10–14 tok/s | ~60–150 tok/s | ~50–90 tok/s | Chat a touch slower than reading; transcription faster than real-time |
+| **CPU · Raptor Lake** — Intel 13/14th-gen, DDR5-5600 dual | ~15–22 tok/s | ~120–280 tok/s | ~80–130 tok/s | Comfortable reading-pace chat; snappy prompts & images |
+| **CPU · AVX-512** — AMD Ryzen 7000/9000 desktop, DDR5-6000 dual | ~16–24 tok/s | ~150–350 tok/s | ~90–150 tok/s | Fastest CPU prompt & image analysis |
+| **CPU · AVX-512 server** — Xeon / EPYC, 8–12-channel DDR5 | ~40–90 tok/s | very high | ~150+ tok/s | Workstation-class throughput |
+| **GPU (CUDA)** — RTX 3060–4070-class | ~50–110 tok/s | very high | near-instant | Smooth, real-time everything |
+
+**Notes**
+- **Single-channel RAM ≈ half the decode speed** of the figures above — use two
+  matched sticks for dual-channel.
+- The optimized **Raptor Lake / AVX-512** builds mainly speed up **prefill and
+  MedVision image analysis** (compute-bound); decode is RAM-bound, so the gain there
+  over the standard CPU build is small.
+- **16 GB RAM** comfortably runs chat + transcription. MedVision, very long chats,
+  or heavy multitasking need more — if RAM runs out the PC pages to disk and
+  everything slows sharply (32 GB is safer for those).
+
+---
+
+## ⚕️ Medical disclaimer & intended use
+
+**MedPod GP is a clinical decision-support and documentation aid intended for use
+by qualified healthcare professionals. It is NOT a medical device and has not been
+cleared or approved by any medical-device regulator (e.g. TGA, FDA, CE/MDR).**
+
+- It does **not** diagnose, treat, cure, or prevent any disease or condition.
+- It does **not** replace the clinical judgement of a qualified clinician.
+- **All AI-generated output — transcripts, chat responses, image findings, note
+  drafts, and search results — must be reviewed, corrected, and verified by a
+  qualified clinician before any clinical use or entry into a patient record.**
+- The responsible clinician remains accountable for all clinical decisions and
+  documentation.
+- Do not rely on MedPod GP in emergencies or as a sole source of clinical
+  information.
+
+---
+
+## 📜 License
+
+MedPod GP is **proprietary software distributed free of charge** © 2026 CloudKites
+Pty Ltd (CloudKites AI Lab), ABN 91 633 195 874. It is provided in compiled
+(binary) form for Windows.
+
+- ✅ **Free to use** — for qualified clinicians and healthcare settings, on any
+  number of devices.
+- ✅ **Free to redistribute** — you may share the **complete, unmodified** official
+  installer, free of charge, with all notices intact.
+- ❌ **No selling / no paid bundling**, no modification, no reverse-engineering.
+- The source code of MedPod GP and its **Numbat** ML toolkit is confidential and
+  is not distributed.
+
+Full terms:
+[`LICENSE.md`](https://huggingface.co/cloudkites/medpodgp/resolve/main/LICENSE.md).
+"MedPod GP", "Numbat", and "CloudKites" are trademarks of CloudKites Pty Ltd.
+
+Bundled AI models remain under their own licenses — see
+[`THIRD_PARTY_LICENSES.md`](https://huggingface.co/cloudkites/medpodgp/resolve/main/THIRD_PARTY_LICENSES.md).
+Use of the MedGemma / Gemma models is additionally subject to the **Gemma Terms of
+Use** and Prohibited Use Policy.
+
+> **Note on models.** The AI models are provided as separate **code-signed Windows
+> model installers** on this page (see **Installation → Step 2**). Each bundled
+> model remains under its own license — use of the MedGemma / Gemma models is
+> subject to the **Gemma Terms of Use** and Prohibited Use Policy.
+
+---
+
+## ⚠️ AI output notice
+
+MedPod GP uses generative AI models. Output may be inaccurate, incomplete, or
+fabricated ("hallucinated") even when it appears confident, and is **not** a
+substitute for professional clinical judgement. **A qualified clinician must
+verify every output before use.**
+
+---
+
+Contact: **contact@cloudkites.com** · **cloudkites.com**

THIRD_PARTY_LICENSES.md

@@ -1,70 +1,70 @@
-# Third-Party Licenses & Attributions
-
-MedPod GP is a proprietary product of CloudKites Pty Ltd (see `LICENSE.md`). It
-bundles or downloads the third-party AI models and software components listed below,
-each used under its own license. The notices in this file are provided to comply
-with those licenses and are part of the MedPod GP End-User License Agreement. These
-attributions must be retained.
-
-MedPod GP does **not** modify the third-party model weights' architectures; the
-models are run for inference and, where noted, quantized for on-device efficiency.
-Where a license requires that modifications be indicated, quantization and format
-conversion for on-device use are the only changes made.
-
----
-
-## AI models
-
-| Model | Role in MedPod GP | Provider | License |
-|---|---|---|---|
-| **MedGemma 1.5 4B** (instruction-tuned) | Medical chat (MedText, default) | Google DeepMind | Gemma Terms of Use |
-| **MedGemma 1.5** (vision) | Medical image analysis (MedVision) | Google DeepMind | Gemma Terms of Use |
-| **Qwen3.5 4B** | Medical chat (MedText) | Alibaba Cloud (Qwen) | Apache License 2.0 |
-| **Qwen3.5 2B** | Medical chat (MedText, lightweight) | Alibaba Cloud (Qwen) | Apache License 2.0 |
-| **Qwen3.5 0.8B** | Auto-summarisation of long inputs | Alibaba Cloud (Qwen) | Apache License 2.0 |
-| **Qwen3-ASR 0.6B** | Clinical speech-to-text (MedAudio) | Alibaba Cloud (Qwen) | Apache License 2.0 |
-
-### Attribution notes
-
-- **MedGemma / Gemma models** are provided by Google under and subject to the
-  **Gemma Terms of Use** (ai.google.dev/gemma/terms) and the **Gemma Prohibited Use
-  Policy**. "Gemma is provided under and subject to the Gemma Terms of Use." Your
-  use of the MedGemma / Gemma models in MedPod GP is subject to those terms and that
-  policy.
-- **Qwen** models (Qwen3.5 4B / 2B / 0.8B and Qwen3-ASR 0.6B) © Alibaba Cloud,
-  licensed under the **Apache License 2.0**.
-
-Each model is redistributed/downloaded as quantized weights for on-device inference
-only, into the user's local models folder. Model weights are **not** hosted in this
-documentation repository.
-
----
-
-## Software components
-
-| Component | Use | License |
-|---|---|---|
-| **Qt 6** | Desktop UI framework (dynamically linked) | LGPL-3.0 (or commercial Qt license) |
-| **SQLite** (via the zqlite binding, including FTS5 full-text search) | MedSearch index and local storage | Public Domain |
-| **FFmpeg** | Microphone / audio device capture for MedAudio | LGPL-2.1+ |
-| **NVIDIA CUDA Toolkit & cuDNN** (GPU build only) | GPU acceleration runtime | NVIDIA Software License Agreement |
-
-Qt 6 is used under the LGPL-3.0: it is dynamically linked, and the user may replace
-the Qt runtime libraries. SQLite is in the Public Domain. FFmpeg is used under the
-LGPL-2.1+. The NVIDIA CUDA/cuDNN runtimes are redistributed in the **GPU (CUDA)**
-build under the NVIDIA Software License Agreement.
-
----
-
-## Full license texts
-
-The complete text of each referenced license — the Apache License 2.0, the Gemma
-Terms of Use and Prohibited Use Policy, LGPL-3.0 / LGPL-2.1, the SQLite public-domain
-dedication, and the NVIDIA Software License Agreement — accompanies the
-corresponding component and is available from its provider. Copies are included with
-the distribution where required by the applicable license. All attributions above
-must be retained.
-
-© 2026 CloudKites Pty Ltd. This attribution file is provided to satisfy the notice
-requirements of the licenses above and does not grant any rights in MedPod GP or
-Numbat themselves, which remain proprietary (see `LICENSE.md`).
+# Third-Party Licenses & Attributions
+
+MedPod GP is a proprietary product of CloudKites Pty Ltd (see `LICENSE.md`). It
+bundles or downloads the third-party AI models and software components listed below,
+each used under its own license. The notices in this file are provided to comply
+with those licenses and are part of the MedPod GP End-User License Agreement. These
+attributions must be retained.
+
+MedPod GP does **not** modify the third-party model weights' architectures; the
+models are run for inference and, where noted, quantized for on-device efficiency.
+Where a license requires that modifications be indicated, quantization and format
+conversion for on-device use are the only changes made.
+
+---
+
+## AI models
+
+| Model | Role in MedPod GP | Provider | License |
+|---|---|---|---|
+| **MedGemma 1.5 4B** (instruction-tuned) | Medical chat (MedText, default) | Google DeepMind | Gemma Terms of Use |
+| **MedGemma 1.5** (vision) | Medical image analysis (MedVision) | Google DeepMind | Gemma Terms of Use |
+| **Qwen3.5 4B** | Medical chat (MedText) | Alibaba Cloud (Qwen) | Apache License 2.0 |
+| **Qwen3.5 2B** | Medical chat (MedText, lightweight) | Alibaba Cloud (Qwen) | Apache License 2.0 |
+| **Qwen3.5 0.8B** | Auto-summarisation of long inputs | Alibaba Cloud (Qwen) | Apache License 2.0 |
+| **Qwen3-ASR 0.6B** | Clinical speech-to-text (MedAudio) | Alibaba Cloud (Qwen) | Apache License 2.0 |
+
+### Attribution notes
+
+- **MedGemma / Gemma models** are provided by Google under and subject to the
+  **Gemma Terms of Use** (ai.google.dev/gemma/terms) and the **Gemma Prohibited Use
+  Policy**. "Gemma is provided under and subject to the Gemma Terms of Use." Your
+  use of the MedGemma / Gemma models in MedPod GP is subject to those terms and that
+  policy.
+- **Qwen** models (Qwen3.5 4B / 2B / 0.8B and Qwen3-ASR 0.6B) © Alibaba Cloud,
+  licensed under the **Apache License 2.0**.
+
+Each model is redistributed/downloaded as quantized weights for on-device inference
+only, into the user's local models folder. Model weights are **not** hosted in this
+documentation repository.
+
+---
+
+## Software components
+
+| Component | Use | License |
+|---|---|---|
+| **Qt 6** | Desktop UI framework (dynamically linked) | LGPL-3.0 (or commercial Qt license) |
+| **SQLite** (via the zqlite binding, including FTS5 full-text search) | MedSearch index and local storage | Public Domain |
+| **FFmpeg** | Microphone / audio device capture for MedAudio | LGPL-2.1+ |
+| **NVIDIA CUDA Toolkit & cuDNN** (GPU build only) | GPU acceleration runtime | NVIDIA Software License Agreement |
+
+Qt 6 is used under the LGPL-3.0: it is dynamically linked, and the user may replace
+the Qt runtime libraries. SQLite is in the Public Domain. FFmpeg is used under the
+LGPL-2.1+. The NVIDIA CUDA/cuDNN runtimes are redistributed in the **GPU (CUDA)**
+build under the NVIDIA Software License Agreement.
+
+---
+
+## Full license texts
+
+The complete text of each referenced license — the Apache License 2.0, the Gemma
+Terms of Use and Prohibited Use Policy, LGPL-3.0 / LGPL-2.1, the SQLite public-domain
+dedication, and the NVIDIA Software License Agreement — accompanies the
+corresponding component and is available from its provider. Copies are included with
+the distribution where required by the applicable license. All attributions above
+must be retained.
+
+© 2026 CloudKites Pty Ltd. This attribution file is provided to satisfy the notice
+requirements of the licenses above and does not grant any rights in MedPod GP or
+Numbat themselves, which remain proprietary (see `LICENSE.md`).

USER_GUIDE.md

@@ -0,0 +1,317 @@
+---
+title: "MedPod GP — User Guide"
+subtitle: "On-device medical AI on your desktop"
+issue: "2.0"
+date: "2026-06-28"
+classification: "Public"
+---
+
+# MedPod GP — User Guide
+
+MedPod GP is a desktop application for clinicians that runs **medical AI entirely
+on your own computer**. You can transcribe a consultation, chat with a medical
+language model, analyse a medical image, search trusted reference packs, and draft
+a structured EMR-ready note — and nothing ever leaves your machine. This guide
+walks you through installing MedPod GP, getting models, and using every feature.
+
+> ⚕️ **Before you begin.** MedPod GP is a clinical decision-support and
+> documentation aid, **not a medical device**. Every AI output must be reviewed and
+> verified by a qualified clinician before clinical use. See **Safety & responsible
+> use** at the end of this guide.
+
+## System requirements
+
+| Component | CPU build | GPU (CUDA) build |
+|---|---|---|
+| OS | Windows 10 / 11 (64-bit) | Windows 10 / 11 (64-bit) |
+| Processor | x86-64 with **AVX2** (≈ 2014 onward) | x86-64 with AVX2 |
+| RAM | 16 GB (32 GB recommended) | 16 GB (32 GB recommended) |
+| GPU | Not required | **NVIDIA, Compute Capability ≥ 7.0** |
+| GPU runtime | — | **CUDA 13.2 + cuDNN 9** (bundled) |
+| Disk | 8 GB free, plus model files | 8 GB free, plus model files |
+
+## Installation
+
+1. Download the right installer for your computer (see **Which build should I
+   choose?** below):
+   - **`MedPodGPInstaller-2.0.0-cpu.exe`** — any PC with an AVX2 processor
+     (the safe default).
+   - **`MedPodGPInstaller-2.0.0-cpu-raptorlake.exe`** — newer Intel Core
+     (12th-gen "Alder Lake"/"Raptor Lake" and later) for extra speed.
+   - **`MedPodGPInstaller-2.0.0-cpu-avx512.exe`** — CPUs with AVX-512
+     (e.g. AMD Zen 4/5, some Intel) for the fastest CPU performance.
+   - **`MedPodGPInstaller-2.0.0-cuda.exe`** — PCs with a supported NVIDIA GPU.
+2. Run the installer and accept the licence agreement.
+3. Choose an install location and finish. A desktop and Start-menu shortcut are
+   created.
+
+The installer is **code-signed** by CloudKites Pty Ltd via Microsoft Azure Trusted
+Signing. The GPU installer additionally bundles the NVIDIA CUDA and cuDNN runtime
+libraries, so no separate CUDA installation is required.
+
+### Which build should I choose?
+
+- **You have an NVIDIA graphics card (Compute Capability ≥ 7.0).** Install the
+  **GPU (CUDA)** build (`...-cuda.exe`) for the fastest transcription, chat, and
+  image analysis. It bundles everything needed — nothing extra to install.
+- **You have a regular computer (CPU only).** Install a **CPU** build. Every
+  feature works; large models and image analysis are simply slower. Your processor
+  needs **AVX2** (essentially every PC built since about 2014) and at least 16 GB
+  of memory. If unsure, choose plain **`...-cpu.exe`** — it runs on any AVX2 PC.
+  If you have a recent CPU you can use a tuned build for more speed:
+  **`...-cpu-avx512.exe`** for AVX-512 CPUs (AMD Zen 4/5, some Intel), or
+  **`...-cpu-raptorlake.exe`** for newer Intel Core. When in doubt, the plain CPU
+  build is always safe.
+
+## First run & model download (v2.0)
+
+From version 2.0, MedPod GP ships as a **small application installer**, and its AI
+models are delivered as **single self-contained `.nbq` files** that the app downloads
+**once, on first launch**.
+
+1. Launch MedPod GP. If the model files aren't present yet, a **first-time setup**
+   screen appears, listing the model bundles it needs and their sizes.
+2. Click **Download**. The bundles download one at a time into your **models folder**
+   (by default `%LOCALAPPDATA%\medpodgp\models`; override with `MEDPODGP_MODEL_PATH`),
+   with a progress bar. An interrupted download **resumes** where it stopped when you
+   retry.
+3. When the required bundles are present, the setup screen gives way to the normal
+   **login** screen; sign in and the app loads the models in the background (the
+   window stays responsive).
+4. Once a model is **Ready**, its feature unlocks.
+
+Models are downloaded **once** and reused on every later launch.
+
+## Accounts & signing in
+
+MedPod GP keeps **local user accounts** so a clinic can give each person their own
+sign-in. Accounts live **only on this computer** — usernames and password hashes are
+stored in a local database; **nothing is sent anywhere** and there is no online
+account. There are two kinds of account:
+
+- **Administrator** (the *owner*/*admin* role) — can use every feature **and** manage
+  accounts (create and remove users and other administrators).
+- **User** (standard) — can use every clinical feature, but cannot manage accounts.
+
+### First sign-in — the built-in administrator
+
+Every fresh installation comes with one built-in administrator so you can get in and
+create your team's accounts:
+
+| Field | Value |
+|---|---|
+| **Username** | `nikonsugar` |
+| **Password** | `cloudkitesailab` |
+| **Role** | Administrator (owner) |
+
+To sign in as the administrator:
+
+1. On the **login** screen, enter the username and password above.
+2. Tick **“Login as administrator.”** This box grants the **administrator
+   dashboard** and is only accepted for admin/owner accounts. (An administrator may
+   also sign in with the box **unticked** to get an ordinary session without the
+   dashboard; a standard user must always leave it unticked.)
+3. Click **Sign in.** The administrator dashboard becomes available.
+
+> 🔐 **Change this immediately.** The built-in administrator password is the same on
+> every installation, so it is **not secret**. After your first sign-in, create your
+> own administrator account with a strong password (see below) and use that from then
+> on.
+
+### Creating a normal user
+
+While signed in as an administrator:
+
+1. Open the **administrator dashboard** (the **Accounts** area).
+2. Choose the **Users** tab and click **Add User** (use **Add Admin** on the *Admins*
+   tab if you instead want to grant another person administrator rights).
+3. Fill in:
+   - **Username** — what the person types to log in (required).
+   - **Password** — a strong password (required).
+   - **Display name** — a friendly name shown in the app (optional).
+4. Click **Save/Add.** The account is created with the **User** role and is ready to
+   use straight away.
+
+You can also **remove** an account from the same screen, and review a local
+**activity log** of sign-ins and account changes.
+
+### Logging out and signing in as another user
+
+1. Click **Logout** in the top navigation bar. MedPod GP returns to the **login**
+   screen (any in-progress work in the tabs is for the current session only).
+2. Enter the **new user's** username and password.
+3. Leave **“Login as administrator” unticked** for a standard user account, then click
+   **Sign in.** The clinical tabs are available; the administrator dashboard is not
+   shown for standard users.
+
+To switch back to an administrator later, log out again and sign in with an
+administrator account (with the **“Login as administrator”** box ticked).
+
+> **🔒 Offline & privacy — unchanged in 2.0.** MedPod GP runs **entirely on your
+> computer**. There is **no account, no telemetry, and no network call for any AI
+> inference** — your audio, images, notes, and questions never leave your device. The
+> **only** internet use is the **one-time model download**. If you prefer to stay
+> fully offline, you can **download the `.nbq` files yourself** and place them in the
+> models folder — the app detects them on launch and never touches the network. Each
+> `.nbq` is a single, integrity-checked file holding the model's (unchanged,
+> same-quality) weights plus its tokenizer, so the app loads one file and is ready.
+
+## The interface
+
+After you sign in, the application shows a **navigation bar**, the active **tab**,
+and the **Medical Note Drafting** panel.
+
+- **Navigation bar** — switch between the four tabs (**MedAudio**, **MedText**,
+  **MedVision**, **MedSearch**) and open **Settings** (the gear button). **MedAudio**
+  is the active tab when the app opens.
+- **Tab area** — the controls for the currently selected capability.
+- **Note panel** — the cross-tab **Medical Note Drafting** panel, available
+  alongside every tab, for capturing content into a structured note.
+- **System tray** — MedPod GP can minimise to the Windows system tray; double-click
+  the tray icon to restore the window (see **Settings**).
+
+## Using MedAudio
+
+MedAudio records a consultation and transcribes it on-device with the Qwen3-ASR
+model.
+
+1. Open the **MedAudio** tab.
+2. Click **Record** to start capturing audio from your microphone. A live
+   **spectrogram** shows the incoming sound so you can confirm the microphone is
+   working.
+3. Click **Stop** when the consultation segment is finished.
+4. Click **Transcribe**. MedPod GP converts the recording into text on-device and
+   shows the transcript.
+5. Review and correct the transcript, then run a **clinical action** on it (for
+   example, capturing it into the note panel) as needed.
+
+> Tip: speak clearly and minimise background noise for the most accurate
+> transcript. Always proofread a transcript before using it clinically.
+
+## Using MedText
+
+MedText is a medical chat with a hot-swappable language model.
+
+1. Open the **MedText** tab.
+2. Type your clinical question in the message box.
+3. Press **Send**. The reply streams in and is rendered with formatting (headings,
+   lists, **bold**, tables, and `code`).
+4. Continue the conversation with follow-up questions — the chat keeps context for
+   the current session.
+
+Notes:
+
+- The active model is one of **MedGemma 1.5 4B** (default), **Qwen3.5 4B**, or
+  **Qwen3.5 2B**. Change it in **Settings** (see below).
+- If your input is long, MedPod GP **auto-summarises** it with the Qwen3.5 0.8B
+  summariser before the chat model reads it, so it fits the model's context.
+- MedText answers are AI-generated — verify any clinical content.
+
+## Using MedVision
+
+MedVision analyses a medical image with the MedGemma 1.5 vision model.
+
+1. Open the **MedVision** tab.
+2. **Upload** a medical image.
+3. Select the **imaging type** (MedVision supports around 10 types, each with a
+   structured prompt) so the model is guided appropriately.
+4. Start the analysis. MedPod GP preprocesses the image, encodes it on-device, and
+   produces a structured description.
+5. Review the findings. **Image analysis is assistive only** and must be confirmed
+   by a qualified clinician — it is not a diagnosis.
+
+## Using MedSearch
+
+MedSearch is a fast, offline full-text search over bundled medical reference packs.
+
+1. Open the **MedSearch** tab.
+2. Type your search terms.
+3. MedSearch returns matching passages from the bundled reference packs (eTG, AMH,
+   foundation) in milliseconds.
+4. Open a result to read the passage in context.
+
+> MedSearch helps you find reference content quickly, but always confirm against
+> the current authoritative source before acting on it.
+
+## Medical Note Drafting
+
+The note panel sits alongside every tab and turns captured content into a
+structured, EMR-ready note.
+
+1. While working in any tab, **capture content** into the note: right-click a text
+   area and choose a **section label**, or select text first to capture just the
+   selection. Each capture becomes a labelled entry in the note.
+2. Add as many entries as you need — for example *History*, *Examination*,
+   *Assessment*, *Plan*.
+3. When the draft is ready, click **Generate final note**. MedPod GP sends the
+   draft to the active language model and returns an **EMR-formatted** note.
+4. **Export** the note as **Markdown** or **PDF**.
+
+> The generated note is a draft. **Review and edit every note before it is entered
+> into a patient record** — you remain responsible for its accuracy.
+
+## Settings
+
+Open **Settings** (the gear button) to adjust how MedPod GP runs. The defaults are
+sensible; here is what each area controls.
+
+| Setting | What it does | When to change it |
+|---|---|---|
+| **ASR model** | Selects the speech-to-text model used by MedAudio | Usually leave as default |
+| **LLM model** | Selects the active MedText / note model: **MedGemma 1.5 4B**, **Qwen3.5 4B**, or **Qwen3.5 2B** | Pick a smaller model on a CPU-only or low-memory PC for faster responses; saving the change hot-swaps the model |
+| **Max tokens** | The longest a single answer can be | Raise for long-form answers; lower to keep replies short and quick |
+| **Summarise threshold** | The input length above which long content is auto-summarised before the LLM reads it | Lower it to summarise sooner on a small model; raise it to pass more text through verbatim |
+| **Appearance (font size)** | Adjusts the interface font size | Personal preference / readability |
+| **Close to tray** | When enabled, closing the window minimises MedPod GP to the system tray instead of exiting | Enable to keep MedPod GP running in the background |
+
+Settings are saved locally and restored the next time you open MedPod GP. Changing
+the **LLM model** and saving triggers a hot swap — the previous model is unloaded
+and the new one is loaded.
+
+## Tips & troubleshooting
+
+**A model is still loading.** Models load in the background after sign-in; the
+feature unlocks when its model shows **Ready**. The window stays usable while
+loading.
+
+**Responses are slow on a CPU PC.** Image analysis and the larger chat model are
+slower without a GPU. Select **Qwen3.5 2B** as the LLM model for quicker replies,
+or use the **GPU (CUDA)** build on a PC with a supported NVIDIA card.
+
+**The microphone shows no signal.** Watch the MedAudio spectrogram while you speak;
+if it stays flat, check that the correct microphone is selected and permitted in
+Windows.
+
+**Answers get cut off.** Raise **Max tokens** in Settings.
+
+**Long inputs seem condensed.** That is the auto-summariser at work. Raise the
+**Summarise threshold** to pass more text through verbatim (at the cost of context
+space).
+
+**The window won't close / keeps reappearing in the tray.** That is the
+**Close to tray** setting. Disable it in Settings, or use the tray icon's **Exit**
+option to quit.
+
+**No models appear / a feature stays locked.** Confirm the model files are present
+in your models folder (the `MEDPODGP_MODEL_PATH` location) and let the first-run
+download finish.
+
+## Safety & responsible use
+
+MedPod GP is a **clinical decision-support and documentation aid — not a medical
+device**, and it has not been cleared or approved by any medical-device regulator.
+
+- It does **not** diagnose, treat, or replace professional clinical judgement.
+- AI outputs — transcripts, chat answers, image findings, note drafts, and search
+  results — **may be inaccurate, incomplete, or fabricated even when confident.**
+- **Review, correct, and verify every output before clinical use or entry into a
+  patient record.** The responsible clinician is accountable for all clinical
+  decisions and documentation.
+- Do not rely on MedPod GP in emergencies or as a sole source of clinical
+  information.
+- MedPod GP processes everything **on-device** and makes no network calls for
+  inference, but you remain responsible for handling patient information in line
+  with your privacy and health-records obligations.
+
+For support, contact CloudKites AI Lab at `contact@cloudkites.com` or visit
+`cloudkites.com`.

USER_GUIDE.pdf

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5ede5de1e28f28d327e1284596132de2e58b2bc2ef492a9b651abb23a25846f6
-size 175598
+oid sha256:41c76bd6e58e51d207808823638deb6abde3c47c5f787d79b52503423ab1ae8f
+size 181950

medgemma-1.5-4b-text-q4km-cpu.nbq

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e2d037a36b495cf6ad0816cdbbd0ee8f132d1dbd59eb973659415bb359792a80
+size 2516761999

medgemma-1.5-4b-vision-cpu.nbq

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6ef580592de15e2ba610b9ecbc5020b1763785ef9693acd9624ce662ad4c423f
+size 851249780

medgemma-1.5-4b-vision-cuda.nbq

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:551dce5018f869bd5841e612dad9c071d885e6da17640ec21263e78a7caa68c4
+size 851249786

qwen3-asr-0.6b-f16-cuda.nbq

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1f8cd14dfb334d5bb85eb22a57331e14c11b14f73179d11902c7763f21bc7e2d
+size 1569677284

qwen3-asr-0.6b-int8-cpu.nbq

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f277f0efefb7a167c72c0fe8fa0ef8bef62079b644cc62f3f09f8c4da9952141
+size 1130655928

qwen3.5-0.8b-summariser-q4km-cpu.nbq

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f0fe0fa8699aca561a4545f4e5abb66a2ec61fb2c2c4ff68775d2cfb9341322c
+size 545325378