deploy to HF

.dockerignore

@@ -7,4 +7,3 @@ __pycache__
 .env.example
 extension
 tests
-flowpilot-vertex-key.json

.github/workflows/deploy.yml

@@ -1,4 +1,4 @@
-name: Deploy Backend
+name: Deploy To Hugging Face Space
 
 on:
   push:
@@ -6,56 +6,26 @@ on:
       - main
   workflow_dispatch:
 
-env:
-  PROJECT_ID: flow-pilot-493104
-  PROJECT_NUMBER: "707525727191"
-  REGION: us-central1
-  SERVICE: flowpilot-backend
-  REPOSITORY: flowpilot
-  IMAGE: backend
-  WIF_PROVIDER: projects/707525727191/locations/global/workloadIdentityPools/github-pool/providers/github-provider
-  SERVICE_ACCOUNT: github-deployer@flow-pilot-493104.iam.gserviceaccount.com
-
 jobs:
   deploy:
     runs-on: ubuntu-latest
     permissions:
       contents: read
-      id-token: write
 
     steps:
       - name: Checkout
         uses: actions/checkout@v4
-
-      - name: Authenticate to Google Cloud
-        uses: google-github-actions/auth@v3
         with:
-          workload_identity_provider: ${{ env.WIF_PROVIDER }}
-          service_account: ${{ env.SERVICE_ACCOUNT }}
-
-      - name: Setup gcloud
-        uses: google-github-actions/setup-gcloud@v2
-
-      - name: Configure Docker auth
-        run: gcloud auth configure-docker ${{ env.REGION }}-docker.pkg.dev --quiet
+          fetch-depth: 0
 
-      - name: Build image
+      - name: Push to Hugging Face Space
+        env:
+          HF_TOKEN: ${{ secrets.HF_TOKEN }}
+          HF_SPACE_REPO: ${{ secrets.HF_SPACE_REPO }}
         run: |
-          docker build \
-            -t ${{ env.REGION }}-docker.pkg.dev/${{ env.PROJECT_ID }}/${{ env.REPOSITORY }}/${{ env.IMAGE }}:${{ github.sha }} \
-            .
-
-      - name: Push image
-        run: |
-          docker push \
-            ${{ env.REGION }}-docker.pkg.dev/${{ env.PROJECT_ID }}/${{ env.REPOSITORY }}/${{ env.IMAGE }}:${{ github.sha }}
-
-      - name: Deploy to Cloud Run
-        uses: google-github-actions/deploy-cloudrun@v3
-        with:
-          service: ${{ env.SERVICE }}
-          region: ${{ env.REGION }}
-          image: ${{ env.REGION }}-docker.pkg.dev/${{ env.PROJECT_ID }}/${{ env.REPOSITORY }}/${{ env.IMAGE }}:${{ github.sha }}
-          flags: >-
-            --allow-unauthenticated
-            --set-env-vars=AI_PROVIDER=vertex_ai,VERTEX_PROJECT_ID=${{ env.PROJECT_ID }},VERTEX_LOCATION=${{ env.REGION }},VERTEX_MODEL=gemini-2.5-flash
+          test -n "$HF_TOKEN"
+          test -n "$HF_SPACE_REPO"
+          git config user.name "github-actions[bot]"
+          git config user.email "41898282+github-actions[bot]@users.noreply.github.com"
+          git remote add hf "https://oauth2:${HF_TOKEN}@huggingface.co/spaces/${HF_SPACE_REPO}"
+          git push hf HEAD:main --force

.gitignore

@@ -18,6 +18,5 @@ build/
 .idea/
 .vscode/
 
-flowpilot-vertex-key.json
 *.pem
 *.key

Dockerfile

@@ -2,7 +2,7 @@ FROM python:3.12-slim
 
 ENV PYTHONDONTWRITEBYTECODE=1
 ENV PYTHONUNBUFFERED=1
-ENV PORT=8080
+ENV PORT=7860
 
 WORKDIR /app
 
@@ -11,4 +11,6 @@ RUN pip install --no-cache-dir -r /app/backend/requirements.txt
 
 COPY backend /app/backend
 
-CMD ["sh", "-c", "uvicorn backend.main:app --host 0.0.0.0 --port ${PORT:-8080}"]
+EXPOSE 7860
+
+CMD ["sh", "-c", "uvicorn backend.main:app --host 0.0.0.0 --port ${PORT:-7860}"]

README.md

@@ -1,11 +1,21 @@
+---
+title: FlowPilot
+emoji: "📬"
+colorFrom: blue
+colorTo: green
+sdk: docker
+app_port: 7860
+short_description: Gmail-first small business workflow automation with FastAPI.
+---
+
 # FlowPilot
 
-FlowPilot is a Gmail-first automation layer for small business owners. The current scaffold includes:
+FlowPilot is a Gmail-first automation layer for small business owners. The project includes:
 
 - A FastAPI backend with analysis, workflow suggestion, workflow build, deploy, upload, status, and escalation endpoints.
 - A primitive-based workflow engine that compiles and executes JSON workflows.
-- A Chrome extension scaffold that injects a sidebar into Gmail and walks through the onboarding-to-live journey.
-- Lightweight in-memory storage and tests for the core owner flow.
+- A Chrome extension scaffold that injects a sidebar into Gmail and walks through the onboarding flow.
+- Lightweight demo storage and tests for the core owner flow.
 
 ## Project Structure
 
@@ -13,67 +23,79 @@ FlowPilot is a Gmail-first automation layer for small business owners. The curre
 backend/
 extension/
 tests/
+Dockerfile
+README.md
 ```
 
-## Backend
+## Local Development
 
-Run locally:
+Install dependencies and run the backend:
 
 ```bash
 pip install -r backend/requirements.txt
 uvicorn backend.main:app --reload
 ```
 
-Create a `.env` in the repo root first. FlowPilot is now wired to read Vertex AI settings from that file.
+Create a `.env` in the repo root first:
 
 ```bash
-cp .env.example .env
+nano .env
 ```
 
-## Extension
+Set these values when you want live Groq responses:
+
+```bash
+AI_PROVIDER=groq
+GROQ_API_KEY=your_groq_key
+GROQ_MODEL=llama-3.3-70b-versatile
+GROQ_BASE_URL=https://api.groq.com/openai/v1
+GROQ_TIMEOUT_SECONDS=8
+```
 
-Load `extension/` as an unpacked Chrome extension. The sidebar injects into Gmail and points to `http://localhost:8000/api`.
+If `GROQ_API_KEY` is blank, FlowPilot falls back to local deterministic logic so the app and tests still work.
 
-## Deploying
+## Hugging Face Space
 
-The repo includes a Cloud Run GitHub Actions workflow at [`.github/workflows/deploy.yml`](/Users/sivasankernp/Desktop/flow-pilot/.github/workflows/deploy.yml:1).
+This repo is ready for a Docker Space. The container listens on port `7860`, and the metadata at the top of this README tells Hugging Face to treat it as a Docker app.
 
-Before the action will work, finish these one-time Google Cloud steps:
+Typical flow:
 
 ```bash
-gcloud iam workload-identity-pools create github-pool \
-  --project=flow-pilot-493104 \
-  --location=global \
-  --display-name="GitHub Pool"
+git clone https://huggingface.co/spaces/technophyle/flow-pilot
+cd flow-pilot
 ```
 
-```bash
-gcloud iam workload-identity-pools providers create-oidc github-provider \
-  --project=flow-pilot-493104 \
-  --location=global \
-  --workload-identity-pool=github-pool \
-  --display-name="GitHub Provider" \
-  --issuer-uri="https://token.actions.githubusercontent.com" \
-  --attribute-mapping="google.subject=assertion.sub,attribute.repository=assertion.repository,attribute.actor=assertion.actor"
-```
+Copy this project into the cloned Space repo, then push:
 
 ```bash
-gcloud iam service-accounts add-iam-policy-binding \
-  github-deployer@flow-pilot-493104.iam.gserviceaccount.com \
-  --project=flow-pilot-493104 \
-  --role="roles/iam.workloadIdentityUser" \
-  --member="principalSet://iam.googleapis.com/projects/707525727191/locations/global/workloadIdentityPools/github-pool/attribute.repository/DevelopedBy-Siva/flow-pilot"
+git add .
+git commit -m "Switch FlowPilot to Hugging Face Space with Groq"
+git push
 ```
 
-The workflow builds the Docker image, pushes it to Artifact Registry, and deploys the FastAPI backend to Cloud Run.
+In the Hugging Face Space settings, add these secrets:
+
+- `GROQ_API_KEY`
+
+Optional Space variables:
+
+- `AI_PROVIDER=groq`
+- `GROQ_MODEL=llama-3.3-70b-versatile`
+- `GROQ_BASE_URL=https://api.groq.com/openai/v1`
+- `GROQ_TIMEOUT_SECONDS=8`
+- `ANALYZE_WITH_AI=false`
+
+For GitHub Actions auto-deploys, add these repository settings:
+
+- Secret: `HF_TOKEN`
+- Variable: `HF_SPACE_REPO=technophyle/flow-pilot`
+
+## Extension
+
+Load `extension/` as an unpacked Chrome extension. The sidebar currently points to `http://localhost:8000/api` for local work, so update that base URL when you connect it to your hosted Space backend.
 
 ## Notes
 
-- Backend storage is currently in-memory for demo speed; the folder layout is ready for SQLite migration work.
-- AI services are still mocked for workflow behavior, but config is now set up for Vertex AI credentials and model selection.
-- Fill in your real Vertex values in `.env`, especially `VERTEX_PROJECT_ID` and `VERTEX_LOCATION`.
-- If you authenticate with `gcloud auth application-default login`, leave `GOOGLE_APPLICATION_CREDENTIALS` blank.
-- Only set `GOOGLE_APPLICATION_CREDENTIALS` when you have a real service account JSON path available.
-- The default model is `gemini-2.5-flash` for better latency and cost during iteration.
-- The backend now attempts live Vertex AI calls through the `google-genai` SDK when credentials are configured.
-- If the SDK is missing or `VERTEX_PROJECT_ID` is still a placeholder, FlowPilot falls back to local deterministic mock logic so tests and scaffolding still work.
+- Backend storage is currently lightweight demo storage suitable for iteration.
+- Groq is used for live AI calls through its OpenAI-compatible chat completions API.
+- The `/api/analyze` endpoint can still stay fast and deterministic when `ANALYZE_WITH_AI=false`.

backend/ai/analyzer.py

@@ -1,13 +1,18 @@
 from __future__ import annotations
 
-from backend.ai.client import render_prompt, vertex_client
+from backend.ai.client import llm_client, render_prompt
+from backend.config import get_settings
 from backend.ai.prompts import ANALYZE_PROMPT, CUSTOM_TASK_PROMPT
 from backend.models.schemas import TaskCategory
 
 
 def analyze_business_description(description: str) -> dict:
-    if vertex_client.is_ready():
-        return vertex_client.generate_json(render_prompt(ANALYZE_PROMPT, description=description))
+    settings = get_settings()
+    if settings.analyze_with_ai and llm_client.is_ready():
+        try:
+            return llm_client.generate_json(render_prompt(ANALYZE_PROMPT, description=description))
+        except Exception:
+            pass
 
     lower = description.lower()
     fully_automatable = []
@@ -71,8 +76,8 @@ def analyze_business_description(description: str) -> dict:
 
 
 def analyze_custom_task(business_description: str, existing_workflows: list[dict], custom_task: str) -> dict:
-    if vertex_client.is_ready():
-        return vertex_client.generate_json(
+    if llm_client.is_ready():
+        return llm_client.generate_json(
             render_prompt(
                 CUSTOM_TASK_PROMPT,
                 business_description=business_description,

backend/ai/classifier.py

@@ -1,10 +1,10 @@
-from backend.ai.client import render_prompt, vertex_client
+from backend.ai.client import llm_client, render_prompt
 from backend.ai.prompts import CLASSIFY_PROMPT
 
 
 def classify_email(from_email: str, subject: str, body: str, business_context: str) -> dict:
-    if vertex_client.is_ready():
-        return vertex_client.generate_json(
+    if llm_client.is_ready():
+        return llm_client.generate_json(
             render_prompt(
                 CLASSIFY_PROMPT,
                 categories="order, availability_inquiry, complaint, greeting, other",

backend/ai/client.py

@@ -1,60 +1,55 @@
 from __future__ import annotations
 
 import json
-import os
 from typing import Any
 
-from backend.config import get_settings
+import httpx
 
-try:
-    from google import genai
-except ImportError:  # pragma: no cover - depends on optional package
-    genai = None
+from backend.config import get_settings
 
 
-class VertexAIClient:
+class GroqAIClient:
     def __init__(self) -> None:
         self.settings = get_settings()
-        self._client = None
 
     def generate_json(self, prompt: str) -> dict[str, Any]:
         text = self.generate_text(prompt)
         return _extract_json_object(text)
 
     def generate_text(self, prompt: str) -> str:
-        client = self._get_client()
-        response = client.models.generate_content(
-            model=self.settings.vertex_model,
-            contents=prompt,
+        settings = get_settings()
+        response = httpx.post(
+            f"{settings.groq_base_url.rstrip('/')}/chat/completions",
+            headers={
+                "Authorization": f"Bearer {settings.groq_api_key}",
+                "Content-Type": "application/json",
+            },
+            json={
+                "model": settings.groq_model,
+                "messages": [
+                    {
+                        "role": "system",
+                        "content": (
+                            "You are a precise backend assistant. "
+                            "Follow the prompt exactly and return valid JSON when requested."
+                        ),
+                    },
+                    {"role": "user", "content": prompt},
+                ],
+                "temperature": 0.2,
+            },
+            timeout=settings.groq_timeout_seconds,
         )
-        text = getattr(response, "text", None)
-        if not text:
-            raise RuntimeError("Vertex AI returned an empty response")
-        return text.strip()
+        response.raise_for_status()
+        payload = response.json()
+        try:
+            return payload["choices"][0]["message"]["content"].strip()
+        except (KeyError, IndexError, AttributeError) as exc:
+            raise RuntimeError("Groq returned an unexpected response shape") from exc
 
     def is_ready(self) -> bool:
-        placeholders = {"your-gcp-project-id"}
-        if self.settings.ai_provider != "vertex_ai":
-            return False
-        if genai is None:
-            return False
-        if self.settings.vertex_project_id in placeholders:
-            return False
-        credentials_path = self.settings.google_application_credentials.strip()
-        if credentials_path and credentials_path != "/absolute/path/to/service-account.json":
-            return os.path.exists(credentials_path)
-        return True
-
-    def _get_client(self):
-        if genai is None:
-            raise RuntimeError("google-genai is not installed")
-        if self._client is None:
-            self._client = genai.Client(
-                vertexai=True,
-                project=self.settings.vertex_project_id,
-                location=self.settings.vertex_location,
-            )
-        return self._client
+        settings = get_settings()
+        return settings.ai_provider == "groq" and bool(settings.groq_api_key.strip())
 
 
 def _extract_json_object(text: str) -> dict[str, Any]:
@@ -70,7 +65,7 @@ def _extract_json_object(text: str) -> dict[str, Any]:
     return json.loads(cleaned[start : end + 1])
 
 
-vertex_client = VertexAIClient()
+llm_client = GroqAIClient()
 
 
 def render_prompt(template: str, **kwargs: Any) -> str:

backend/ai/composer.py

@@ -1,12 +1,12 @@
 import json
 
-from backend.ai.client import render_prompt, vertex_client
+from backend.ai.client import llm_client, render_prompt
 from backend.ai.prompts import COMPOSE_PROMPT
 
 
 def compose_reply(context: dict, tone: str = "friendly") -> str:
-    if vertex_client.is_ready():
-        return vertex_client.generate_text(
+    if llm_client.is_ready():
+        return llm_client.generate_text(
             render_prompt(
                 COMPOSE_PROMPT,
                 business_name=context.get("business_name", "FlowPilot customer"),

backend/ai/extractor.py

@@ -1,10 +1,10 @@
-from backend.ai.client import render_prompt, vertex_client
+from backend.ai.client import llm_client, render_prompt
 from backend.ai.prompts import EXTRACT_PROMPT
 
 
 def extract_email_data(from_email: str, subject: str, body: str) -> dict:
-    if vertex_client.is_ready():
-        return vertex_client.generate_json(
+    if llm_client.is_ready():
+        return llm_client.generate_json(
             render_prompt(
                 EXTRACT_PROMPT,
                 from_email=from_email,

backend/ai/workflow_builder.py

@@ -1,12 +1,12 @@
 import json
 
-from backend.ai.client import render_prompt, vertex_client
+from backend.ai.client import llm_client, render_prompt
 from backend.ai.prompts import BUILD_WORKFLOW_PROMPT
 from backend.models.schemas import BuildWorkflowRequest, WorkflowDefinition, WorkflowStep
 
 
 def build_workflow_definition(request: BuildWorkflowRequest, owner: dict) -> dict:
-    if vertex_client.is_ready():
+    if llm_client.is_ready():
         prompt = render_prompt(
             BUILD_WORKFLOW_PROMPT,
             selected_option=json.dumps(request.selected_option, indent=2),
@@ -19,7 +19,7 @@ def build_workflow_definition(request: BuildWorkflowRequest, owner: dict) -> dic
             stock_column="stock",
             uploaded_data_summary=json.dumps(owner.get("uploaded_data_summary", []), indent=2),
         )
-        return vertex_client.generate_json(prompt)
+        return llm_client.generate_json(prompt)
 
     trigger_type = "schedule" if "summary" in request.task_name.lower() else "email_received"
     steps: list[WorkflowStep] = []

backend/ai/workflow_suggester.py

@@ -1,4 +1,4 @@
-from backend.ai.client import render_prompt, vertex_client
+from backend.ai.client import llm_client, render_prompt
 from backend.ai.prompts import SUGGEST_WORKFLOWS_PROMPT
 
 
@@ -9,8 +9,8 @@ def suggest_workflow_options(
     spreadsheet_info: dict,
     uploaded_files: list[dict],
 ) -> dict:
-    if vertex_client.is_ready():
-        return vertex_client.generate_json(
+    if llm_client.is_ready():
+        return llm_client.generate_json(
             render_prompt(
                 SUGGEST_WORKFLOWS_PROMPT,
                 task_name=task_name,

backend/api/routes.py

@@ -159,13 +159,13 @@ def simulate_run(owner_id: str, workflow_id: str, trigger: dict) -> dict:
     return result
 
 
-@router.get("/debug/vertex")
-def debug_vertex() -> dict:
-    from backend.ai.client import vertex_client
+@router.get("/debug/groq")
+def debug_groq() -> dict:
+    from backend.ai.client import llm_client
 
     try:
-        text = vertex_client.generate_text("Reply with exactly: Vertex debug ok")
+        text = llm_client.generate_text("Reply with exactly: Groq debug ok")
         return {"status": "ok", "response": text}
     except Exception as exc:
-        logger.exception("Vertex debug call failed")
-        raise HTTPException(status_code=500, detail=f"Vertex debug failed: {exc}") from exc
+        logger.exception("Groq debug call failed")
+        raise HTTPException(status_code=500, detail=f"Groq debug failed: {exc}") from exc

backend/config.py

@@ -12,15 +12,13 @@ class Settings(BaseModel):
     api_prefix: str = "/api"
     database_url: str = "sqlite:///./flowpilot.db"
     allow_origins: list[str] = Field(default_factory=lambda: _split_csv(os.getenv("ALLOW_ORIGINS", "*")))
-    ai_provider: str = os.getenv("AI_PROVIDER", "vertex_ai")
+    ai_provider: str = os.getenv("AI_PROVIDER", "groq")
+    analyze_with_ai: bool = os.getenv("ANALYZE_WITH_AI", "false").lower() == "true"
     gmail_poll_seconds: int = int(os.getenv("GMAIL_POLL_SECONDS", "30"))
-    vertex_project_id: str = os.getenv("VERTEX_PROJECT_ID", "your-gcp-project-id")
-    vertex_location: str = os.getenv("VERTEX_LOCATION", "us-central1")
-    vertex_model: str = os.getenv("VERTEX_MODEL", "gemini-2.5-flash")
-    google_application_credentials: str = os.getenv(
-        "GOOGLE_APPLICATION_CREDENTIALS",
-        "",
-    )
+    groq_api_key: str = os.getenv("GROQ_API_KEY", "")
+    groq_model: str = os.getenv("GROQ_MODEL", "llama-3.3-70b-versatile")
+    groq_base_url: str = os.getenv("GROQ_BASE_URL", "https://api.groq.com/openai/v1")
+    groq_timeout_seconds: float = float(os.getenv("GROQ_TIMEOUT_SECONDS", "8"))
 
 
 def _split_csv(value: str) -> list[str]:

backend/requirements.txt

@@ -4,4 +4,3 @@ pydantic==2.11.7
 pytest==8.4.1
 httpx==0.28.1
 python-dotenv==1.1.1
-google-genai==1.30.0