Preparar deploy no Hugging Face Spaces.

Adiciona Dockerfile na porta 7860, caches em /tmp e guia de deploy no README.

Co-authored-by: Cursor <cursoragent@cursor.com>

.dockerignore

@@ -0,0 +1,13 @@
+.git
+.gitignore
+.env
+.env.*
+chroma_db/
+__pycache__/
+*.py[cod]
+.venv/
+venv/
+.pytest_cache/
+.ruff_cache/
+*.md
+!README.md

Dockerfile

@@ -0,0 +1,22 @@
+FROM python:3.11-slim
+
+WORKDIR /app
+
+# HF Spaces roda como usuário não-root; caches devem ir para /tmp.
+ENV PORT=7860 \
+    CHROMA_PERSIST_DIR=/tmp/chroma_db \
+    HF_HOME=/tmp/hf_home \
+    TRANSFORMERS_CACHE=/tmp/hf_home \
+    SENTENCE_TRANSFORMERS_HOME=/tmp/hf_home \
+    XDG_CACHE_HOME=/tmp/cache \
+    PYTHONUNBUFFERED=1 \
+    PYTHONDONTWRITEBYTECODE=1
+
+COPY requirements.txt .
+RUN pip install --no-cache-dir -r requirements.txt
+
+COPY . .
+
+EXPOSE 7860
+
+CMD ["sh", "-c", "uvicorn main:app --host 0.0.0.0 --port ${PORT}"]

README.md

@@ -1,3 +1,13 @@
+---
+title: Agent Memory
+emoji: 🧠
+colorFrom: blue
+colorTo: purple
+sdk: docker
+app_port: 7860
+pinned: false
+---
+
 # Agent Memory — Fase 1 (Baseline)
 
 PoC de **memória observável**: chat à esquerda, armazém de memórias à direita.
@@ -14,11 +24,58 @@ agent_memory/
 ├── models.py         # Pydantic models
 ├── seeds.py          # Memórias seed (falas de clientes fictícios)
 ├── requirements.txt
+├── Dockerfile        # Deploy em Hugging Face Spaces
 ├── .env.example
 └── static/
     └── index.html    # UI do armazém
 ```
 
+## Deploy no Hugging Face Spaces
+
+### 1. Criar o Space
+
+1. Acesse [huggingface.co/new-space](https://huggingface.co/new-space)
+2. **Space name:** ex. `agent-memory`
+3. **SDK:** Docker
+4. **Hardware:** CPU basic (grátis)
+5. **Visibility:** Public ou Private
+
+### 2. Conectar ao GitHub (recomendado)
+
+Na aba **Settings** do Space:
+
+1. **Repository** → conecte `EriqueSouza/agent_memory` (branch `claude_agent` ou `main`)
+2. Faça push deste repositório com o `Dockerfile` e o frontmatter YAML no `README.md`
+
+Ou faça push direto para o git do Space:
+
+```bash
+git remote add space https://huggingface.co/spaces/SEU_USUARIO/agent-memory
+git push space claude_agent:main
+```
+
+### 3. Configurar secrets
+
+Em **Settings → Repository secrets** do Space, adicione:
+
+| Secret | Valor |
+|--------|-------|
+| `OPENROUTER_API_KEY` | sua chave OpenRouter |
+| `OPENROUTER_MODEL` | ex. `nvidia/nemotron-3-super-120b-a12b` |
+| `OPENROUTER_API_URL` | `https://openrouter.ai/api/v1` |
+
+> O Space reinicia automaticamente após salvar os secrets.
+
+### 4. Acessar
+
+URL pública: `https://huggingface.co/spaces/SEU_USUARIO/agent-memory`
+
+A UI abre na raiz (`/`). O build inicial pode levar alguns minutos (download do modelo de embedding do ChromaDB).
+
+**Limitações no tier gratuito:** disco efêmero — memórias criadas no chat somem se o Space reiniciar; os seeds são recarregados automaticamente.
+
+---
+
 ## Setup
 
 ### 1. Criar ambiente

agent.py

@@ -15,6 +15,15 @@ from memory_store import MemoryStore
 
 load_dotenv()
 
+
+def _openrouter_referer() -> str:
+    if host := os.getenv("SPACE_HOST"):
+        return host
+    if space_id := os.getenv("SPACE_ID"):
+        return f"https://huggingface.co/spaces/{space_id}"
+    return "http://localhost:8000"
+
+
 SYSTEM_PROMPT = """\
 You are a customer service assistant for an internet, TV, and telephony provider in Brazil.
 This is a memory-observability PoC: you help the current customer resolve their situation using \
@@ -127,7 +136,7 @@ class Agent:
                 "response_format": {"type": "json_object"},
             },
             default_headers={
-                "HTTP-Referer": "http://localhost:8000",
+                "HTTP-Referer": _openrouter_referer(),
                 "X-Title": "Agent Memory Phase 1",
             },
         )

main.py

@@ -1,3 +1,5 @@
+import os
+
 from fastapi import FastAPI, HTTPException
 from fastapi.staticfiles import StaticFiles
 from fastapi.responses import FileResponse
@@ -9,7 +11,8 @@ from agent import Agent
 from seeds import SEED_MEMORIES
 
 
-memory_store = MemoryStore(persist_dir="./chroma_db")
+PERSIST_DIR = os.getenv("CHROMA_PERSIST_DIR", "./chroma_db")
+memory_store = MemoryStore(persist_dir=PERSIST_DIR)
 agent = Agent(memory_store=memory_store)
 
 
@@ -79,4 +82,6 @@ async def reset_memories():
 
 if __name__ == "__main__":
     import uvicorn
-    uvicorn.run("main:app", host="0.0.0.0", port=8000, reload=True)
+
+    port = int(os.getenv("PORT", "8000"))
+    uvicorn.run("main:app", host="0.0.0.0", port=port, reload=True)