Publish merged LedgerGuard V2 accounting workflow model

README.md

@@ -2,99 +2,100 @@
 license: other
 license_name: blackwoodai-personal-use-v1.0
 base_model: Qwen/Qwen3.6-27B
-tags:
-  - accounting
-  - bookkeeping
-  - tool-calling
-  - finance
-  - reconciliation
-  - draft-only
+library_name: transformers
 pipeline_tag: text-generation
+tags:
+- qwen
+- accounting
+- transaction-classification
+- rule-classification
+- lora-merged
+- non-commercial
 ---
 
 # LedgerGuard-27B
 
-LedgerGuard-27B is an accounting workflow model from BlackwoodAI for draft-only bookkeeping and accounting operations. It is built for the work accountants actually repeat: reconciling transactions, preparing review packets, matching invoices and bills, detecting exceptions, and producing source-linked draft actions for approval.
-
-The public Hugging Face release is for personal evaluation. Commercial licensing, hosted access, and firm-specific deployments are available from BlackwoodAI.
-
-## What It Is Built For
-
-LedgerGuard is designed for accounting workflows where a generic assistant is too loose and a hardcoded rules engine is too brittle.
-
-- Bank reconciliation and transaction categorization
-- Invoice, bill, payment, and deposit matching
-- Draft journal-entry preparation
-- Month-end close review support
-- Sales tax, VAT, and tax-code review workflows
-- Missing receipt, stale feed, duplicate transaction, inactive account, and locked-period detection
-- Source-linked reviewer packets for human approval
-
-## Accounting Software Coverage
-
-LedgerGuard V2 work is connector-shaped around the systems firms and operators already use:
-
-- QuickBooks Online
-- Xero
-- Sage Intacct
-- Oracle NetSuite
-- Zoho Books
-- FreshBooks
-- Odoo Accounting
-- Microsoft Dynamics 365 Business Central
-- SAP Business One
-- Wave-style invoice, receipt, and export workflows
-
-The model is trained to think in accounting software objects: bank transactions, ledger entries, chart-of-accounts records, invoices, bills, payments, vendors, customers, tax codes, dimensions, attachments, and approval states.
-
-## How It Works In A Product
-
-LedgerGuard should sit behind a deterministic workflow layer. The model proposes draft actions; your product validates them before anything touches an accounting system.
-
-Recommended runtime:
-
-1. Normalize records from accounting software, bank feeds, statements, and source documents.
-2. Retrieve the relevant ledger, invoice, bill, payment, account, tax, and attachment context.
-3. Ask LedgerGuard for a draft decision with source references.
-4. Run deterministic checks for period locks, duplicate entries, account validity, tax-code validity, balanced journals, missing support, and unsafe write attempts.
-5. Send the reviewer packet to a human before any posting or sync.
-
-LedgerGuard is not meant to autonomously post to a ledger. It is meant to make the accountant review queue faster, cleaner, and easier to trust.
-
-## Where It Helps
-
-LedgerGuard is strongest where a firm has repeated workflows, firm policies, client-specific patterns, and a need for consistent reviewer-ready outputs.
-
-- CAS and bookkeeping teams handling high transaction volume
-- Firms cleaning up messy bank feeds and unmatched transactions
-- Operators preparing month-end support and close review packets
-- Finance teams that need draft accounting actions with evidence
-- Products adding accounting automation without giving an LLM direct posting power
-
-## Safety Posture
-
-Accounting automation should be conservative. LedgerGuard is designed around draft-only behavior:
-
-- No autonomous posting
-- No silent write-back
-- No sync without approval
-- Source references required for material recommendations
-- Human review for missing, conflicting, stale, duplicated, locked, or policy-sensitive cases
-- Deterministic validators expected in production
-
-## Commercial Access
-
-The public model is available for personal evaluation only. For production use, hosted inference, accounting-software integrations, private deployment, or firm-specific customization, contact BlackwoodAI for commercial terms.
-
-BlackwoodAI can provide a productized deployment path around:
-
-- Hosted model access
-- Connector-specific tool schemas
-- Firm policy customization
-- Approval workflows
-- Audit logs and reviewer packets
-- Private evaluation against your target workflow
+This is a full merged derivative model: Qwen/Qwen3.6-27B with a BlackwoodAI
+accounting software workflow specialist fine-tune merged into the weights.
+
+It is intended for personal, non-commercial research and evaluation only. It is
+not licensed for business use, hosted inference, SaaS, or model-as-a-service.
+
+## Why It Exists
+
+Accounting workflows often break on the small details: exact JSON shape,
+source-linked decisions, conservative review flags, draft-only tool calls, and
+software-specific workflow discipline. LedgerGuard-27B is tuned for that
+specialist layer.
+
+It is designed to be useful where a general model may understand the situation
+but drift on the house style.
+
+## Strengths
+
+- Draft-only accounting software workflow outputs
+- Structured accounting-review JSON
+- Source-linked tool-call behavior
+- Conservative handling of ambiguous payment-processor deposits, stale feeds,
+  missing support, duplicate-risk transactions, and review-required items
+- No autonomous posting/write-back behavior by design
+- Useful as a specialist component beside validators and a senior reviewer model
+
+## Intended Output Style
+
+- choose `auto_ready`, `needs_attention`, or `blocked`
+- emit source-linked JSON for review workflows
+- prepare draft-only tool-call payloads for accounting software workflows
+- avoid write-back/posting actions
+
+## Sanitized Example
+
+Input:
+
+```text
+Platform: cloud accounting system
+Workflow: bank reconciliation
+Context: imported bank transaction has partial supporting evidence
+```
+
+Expected style of behavior:
+
+```json
+{
+  "decision": "needs_attention",
+  "evidence_class": "review_required",
+  "source_refs": [{"type": "bank_transaction", "id": "bank-demo-001"}],
+  "tool_calls": [
+    {
+      "tool": "ledgerguard.validate_draft",
+      "arguments": {
+        "case_id": "demo-001",
+        "draft_only": true,
+        "check_source_refs": true
+      }
+    }
+  ],
+  "review_packet": {
+    "approval_required": true,
+    "draft_only": true,
+    "summary": "Route to accountant review before any posting action."
+  }
+}
+```
+
+Plain English: prepare review-ready draft outputs and keep write/post/sync
+actions behind deterministic validators and human approval.
+
+## Evaluation
+
+This model was evaluated on BlackwoodAI internal, non-public accounting
+workflow tests focused on exact JSON/schema adherence, source-linked decisions,
+draft-only safety, and accounting software tool-call discipline. The evaluation
+data, generation process, baselines, and case counts are intentionally not
+disclosed.
 
 ## License
 
-See the model repository license. The public LedgerGuard-27B release is not licensed for business use, hosted inference, SaaS, or model-as-a-service usage without a separate BlackwoodAI commercial agreement.
+See `LICENSE`. This release is personal/non-commercial only and prohibits
+business use, hosted inference, SaaS, and model-as-a-service usage. The upstream
+base model is Qwen/Qwen3.6-27B and remains under its own upstream license.

model-00002-of-00012.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:0f679e9b69dd50ab7eedca4bad1e0ea485b2f456645aed236952e16969022104
+oid sha256:0d32f602732aab4b5eb4d863b9fa12a07a97a6119351d97f2d3181b78dd39637
 size 4842451920

model-00003-of-00012.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d0616300458f58b17b2f538b3f64fcb3efbd7e5bb85f2958fbf9aa30375e28b9
+oid sha256:eb11b825d2236ecc4f0f93a0c41eed97b08279f27fe967fa96fff8b8fa4ba1ed
 size 4965227944

model-00004-of-00012.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4b167e656d053cc47cbd4e39faab579fd600e58563638579ffde857d64c500ea
+oid sha256:c1e5476709e8a38e504958efde940276d6d500cbae24d0cc93fb4bd0b13d98ca
 size 4912819264

model-00005-of-00012.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:405cd97aedbb7dbe9b5af677d0dd9f4e09e6b64f8e0e8ca1ed18c08f3ee138e7
+oid sha256:d93a255d7d9578b3da03ae8ff6c9626ea940936ed841a9ea880c0f3738857667
 size 4986198544

model-00006-of-00012.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:38930c764031c59fa6dfb4e40c45fe36ba98fb4da6f6617a9c8808534b063f48
+oid sha256:4452230cf2964e831e4ed42167d090ecb1d90cd4bb102a4e08c87fa17fca43e7
 size 4912819320

model-00007-of-00012.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ed98481afd5ac49730e43f66c5e47ec2136e7cb24da1460a79e9fd1ef597ca3e
+oid sha256:b754195dacb679bbe96d329b8f213d896c2d11d7b063575df1d96ca7193f78f4
 size 4932703272

model-00008-of-00012.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2c77ad58e9060ab9aba6603a910fdba6b7d36a30acb50537e1810ab4f206638b
+oid sha256:962f92353bd0b00049c218a6fb25668342ceecb16078bfb71e24abab14df8cc8
 size 4966314576

model-00009-of-00012.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:21e16a6d7bb0a8bfc8fc8e49fc3f321379c7d66df1c44e1b1e1e9ea152dcf2cc
+oid sha256:0ec38757b8b7cc45ed5c6c51559828f5758babb90921a6e245098e91265fbe6c
 size 4964162248

model-00010-of-00012.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:780af9ad474875b8090680435c92b635435d8a6b3fa6eac3621b4611a8779d03
+oid sha256:b7005ebb43a943fe9c2d98bf87838db60816d189975a306e6a0731c50af39925
 size 4933789824

model-00011-of-00012.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:de71128099c539f63a6b41a25469afde49f44efc0d4d1cb0e621120aeafb10c0
+oid sha256:e99a1a95d13aa26b07305a1cf7b9c46baffb4e32423d0ebba5596084f3203ee7
 size 4965228032

model-00012-of-00012.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b86f7fc51338cdadf5a32f74ebc5094c774b796b8528ca12fec3c3d74eca2efa
+oid sha256:c53e2e5066ec8379337e4ab6260e03414c501e4ee44ecffb2f710eeffc844fe5
 size 1867596944