Update README.md

README.md

@@ -1,120 +1,13 @@
-# Trading Game Study AI — Hugging Face Space Rebuild v2
-
-This is a publish-ready FastAPI Docker Space for your Unity trading-game study.
-
-## What this version adds
-
-- **Quant support**: rule-based GMAT-style solver with SymPy support
-- **Non-quant conversation**: lightweight Hugging Face **transformers** generator for natural chat
-- **Retrieval**: local or remote **datasets** loading from your GMAT corpus with embedding search + reranking
-- **Researcher-facing logging endpoints** for Unity WebGL studies
-- **Optional Hugging Face dataset persistence** for uploaded logs
-
-## Main endpoints
-
-- `GET /` browser test page
-- `GET /health`
-- `POST /chat`
-- `POST /log/session/start`
-- `POST /log/event`
-- `POST /log/session/finalize`
-- `GET /research/sessions`
-- `GET /research/session/{session_id}`
-
-## Chat behaviour
-
-### Quant messages
-Quant-style messages are routed into the deterministic solver layer.
-The response is designed to help the participant without directly exposing the final answer.
-
-### Non-quant messages
-General chat, AI-usage questions, gameplay questions, and non-math messages are routed into a lightweight generative model.
-
-Default model:
-- `google/flan-t5-small`
-
-You can swap it through environment variables:
-- `GENERATOR_MODEL`
-- `GENERATOR_TASK`
-
-## Logging architecture
-
-This Space now supports researcher-facing logging so Unity WebGL does not have to rely on browser-local files alone.
-
-### Flow
-1. Unity calls `/log/session/start`
-2. Unity sends repeated `/log/event` payloads
-3. Unity calls `/log/session/finalize`
-4. Researcher reads session bundles from `/research/...`
-
-### Local persistence
-Logs are written into:
-- `logs/sessions/`
-- `logs/events/`
-
-### Hugging Face Hub persistence
-If you set these variables, the Space will also push each log file to a dataset repo:
-
-- `PUSH_LOGS_TO_HUB=1`
-- `HF_TOKEN=...`
-- `LOG_DATASET_REPO_ID=your-username/your-study-logs`
-
-This is the recommended option if you want the logs to persist beyond the container lifecycle.
-
-## Recommended Space secrets
-
-- `HF_TOKEN`
-- `INGEST_API_KEY`
-- `RESEARCH_API_KEY`
-
-## Example Unity payloads
-
-### Start session
-```json
-{
-  "participant_id": "P001",
-  "condition": "high_transparency",
-  "study_id": "trading_ai_study",
-  "game_version": "1.0.0",
-  "metadata": {
-    "platform": "WebGL"
-  }
-}
-```
-
-### Event
-```json
-{
-  "participant_id": "P001",
-  "session_id": "SESSION_ID_HERE",
-  "event_type": "AIMessageSent",
-  "question_index": 8,
-  "turn_index": 2,
-  "payload": {
-    "tone": 0.8,
-    "verbosity": 0.6,
-    "transparency": 0.4,
-    "promptLengthChars": 122,
-    "tokensDisplayedBeforeSend": 32
-  }
-}
-```
-
-### Finalize
-```json
-{
-  "participant_id": "P001",
-  "session_id": "SESSION_ID_HERE",
-  "summary": {
-    "totalAiInteractions": 11,
-    "totalTokensSpent": 216,
-    "questionsWithAiUse": 7
-  }
-}
-```
-
-## Notes
-
-- This build is ready to publish as a Hugging Face **Docker Space**.
-- For a real study, I strongly recommend enabling Hub log persistence or wiring an external database later.
-- The local container filesystem alone should not be treated as the final durable storage layer for participant data.
+---
+title: Trading Game AI V2
+emoji: 🤖
+colorFrom: blue
+colorTo: indigo
+sdk: docker
+app_port: 7860
+pinned: false
+---
+
+# Trading Game AI V2
+
+Docker Space for the Unity trading-game AI backend.