---
title: RFAB Historic Chat Space - GPT 1900
emoji: 🕰️
colorFrom: blue
colorTo: purple
sdk: gradio
sdk_version: 5.49.1
app_file: app.py
pinned: false
models:
- Pclanglais/MonadGPT
---

# RFAB Historic Chat Gradio Space Template

This template exposes the same Gradio API shape used by Talkie-1930:

- `POST /gradio_api/call/v2/_bot_reply`
- `GET /gradio_api/call/_bot_reply/{event_id}`

The Reality Fabricator backend expects `_bot_reply` to accept named params:

- `history`
- `system_prompt`
- `temperature`
- `max_tokens`
- `top_p`
- `top_k`

## Required Variables

Set these in the Space settings:

```bash
RFAB_HISTORIC_MODEL_ID=Pclanglais/MonadGPT
RFAB_HISTORIC_ADAPTER=transformers
DRY_RUN=true
```

For real model inference, set `DRY_RUN=false` and choose suitable hardware.

## Model Presets

MonadGPT:

```bash
RFAB_HISTORIC_MODEL_ID=Pclanglais/MonadGPT
RFAB_HISTORIC_ADAPTER=transformers
RFAB_HISTORIC_MODEL_KWARGS={"torch_dtype":"auto","device_map":"auto"}
```

TimeCapsule:

```bash
RFAB_HISTORIC_MODEL_ID=haykgrigorian/TimeCapsuleLLM-v2-llama-1.2B
RFAB_HISTORIC_ADAPTER=transformers
```

GPT-1900:

```bash
RFAB_HISTORIC_MODEL_ID=mhla/gpt1900-instruct-v3-sft
RFAB_HISTORIC_ADAPTER=nanochat
RFAB_NANOCHAT_MODEL_FILE=model_000075.pt
RFAB_NANOCHAT_META_FILE=meta_000075.json
RFAB_GPU_DURATION_SECONDS=240
```

Mr. Chatterbox:

```bash
RFAB_HISTORIC_MODEL_ID=tventurella/mr_chatterbox_model
RFAB_HISTORIC_ADAPTER=nanochat
RFAB_NANOCHAT_MODEL_TAG=d18
RFAB_NANOCHAT_SOURCE=sft
RFAB_NANOCHAT_MODEL_FILE=model_000117.pt
RFAB_NANOCHAT_META_FILE=meta_000117.json
RFAB_GPU_DURATION_SECONDS=240
```

The `nanochat` adapter lazily downloads the checkpoint, loads `tokenizer/`, `meta_*.json`, and `model_*.pt`, then generates with nanochat special tokens. Keep `RFAB_GPU_DURATION_SECONDS` high on first real smoke because the first request downloads and loads the checkpoint.

The template clones nanochat at runtime instead of installing it with pip because the upstream repo is not currently pip-wheel-installable from GitHub. Optional runtime override:

```bash
RFAB_NANOCHAT_REPO=https://github.com/karpathy/nanochat.git
RFAB_NANOCHAT_REF=1076f97
```

## Hardware

- CPU Basic is free and good for build/API contract testing.
- ZeroGPU requires a personal PRO account or qualifying Team/organization setup.
- The same Space can be upgraded later in Settings or through `huggingface_hub`; hardware changes restart the Space.
- Keep `@spaces.GPU` in `app.py`; it is effect-free outside ZeroGPU and required once ZeroGPU is selected.

## Backend Activation

After the Space is live:

```bash
curl https://<space>.hf.space/gradio_api/info
npm run smoke:historic-hf -- --model <model>
```

Copy the smoke JSON into `HISTORIC_HF_SMOKE_RESULTS_JSON` and set the matching backend URL env var, for example:

```bash
HISTORIC_HF_SPACE_URL_MONAD_GPT=https://rfab-historic-monad-gpt.hf.space
```