36 lines
1.8 KiB
Markdown
36 lines
1.8 KiB
Markdown
# ChatPBC Fine-Tuned Deployment Checklist
|
|
|
|
This checklist ensures that ChatPBC V4 and V3.3 are correctly deployed and accessible via the direct Hugging Face Inference API.
|
|
|
|
## 1. Repository Configuration
|
|
- [ ] Model repositories `chatpbc1/chatpbc-v4` and `chatpbc1/chatpbc-v33` are set to **Public**.
|
|
- [ ] Hugging Face Token has `write` access to these repositories.
|
|
|
|
## 2. API Endpoint Verification
|
|
- [ ] **ChatPBC V4 Endpoint:** `https://api-inference.huggingface.co/models/chatpbc1/chatpbc-v4/v1/chat/completions`
|
|
- [ ] **ChatPBC V3.3 Endpoint:** `https://api-inference.huggingface.co/models/chatpbc1/chatpbc-v33/v1/chat/completions`
|
|
- [ ] Endpoint format follows the OpenAI-compatible direct inference structure.
|
|
|
|
## 3. Code Implementation
|
|
- [ ] `ChatPBC_Conversational_Demo.html` uses the direct endpoints.
|
|
- [ ] `hf_space_app.py` uses the direct endpoints.
|
|
- [ ] The `provider` field is **REMOVED** from all API request bodies.
|
|
- [ ] Authorization header is correctly set to `Bearer [HF_TOKEN]`.
|
|
- [ ] UI reflects "ChatPBC" branding only (no mentions of Llama or other base models).
|
|
|
|
## 4. Security & Deployment
|
|
- [ ] The version of `ChatPBC_Conversational_Demo.html` uploaded to repositories uses the placeholder `YOUR_HF_TOKEN_HERE`.
|
|
- [ ] The version delivered to the user contains the real token.
|
|
- [ ] `hf_space_app.py` is uploaded to both repositories.
|
|
- [ ] Supporting scripts (`chatpbc_webscraper.py`, `chatpbc_file_processor.py`) are included.
|
|
|
|
## 5. Functional Testing
|
|
- [ ] Local HTML file opens in browser and successfully communicates with the model.
|
|
- [ ] CORS errors are absent due to using the `/v1/chat/completions` sub-endpoint.
|
|
- [ ] File uploads and URL inputs are correctly appended to the conversation context.
|
|
- [ ] Model switching between V4 and V3.3 works as expected.
|
|
|
|
---
|
|
**Developer:** Mik Tse Agency
|
|
**Date:** July 26, 2026
|