Paython mcp

package.json

@@ -27,6 +27,7 @@
     "@aws-sdk/client-sts": "^3.921.0",
     "@aws-sdk/credential-provider-sso": "^3.921.0",
     "@openai/agents": "^0.2.1",
+    "@supabase/supabase-js": "^2.81.1",
     "axios": "^1.12.2",
     "chalk": "^5.6.2",
     "cookie-parser": "^1.4.7",

server/agent/api_calls_doc.md

@@ -0,0 +1,102 @@
+# MCP CI/CD Builder – API Usage Guide
+
+This short guide explains how to use the **Wizard Agent API** to generate pipelines, confirm workflows, and commit the final YAML to GitHub. It also highlights the new **confirmation flow** that ensures pipelines are only committed when the user explicitly approves.
+
+---
+
+## 🚀 Overview
+The Wizard Agent provides a multi‑step workflow for generating and committing CI/CD pipelines via API calls:
+
+1. **User requests a pipeline** → Agent generates YAML
+2. **Agent asks for confirmation** → `requires_confirmation: true`
+3. **User confirms** → Agent commits workflow to GitHub
+
+The session must include a valid `mcp_session` cookie for authentication.
+
+---
+
+## 📌 Step 1 — Generate a Pipeline
+Send a prompt describing the pipeline you want.
+
+```bash
+curl -X POST http://localhost:3000/agent/wizard/ai \
+  -H "Content-Type: application/json" \
+  -H "Cookie: mcp_session=YOUR_SESSION_TOKEN" \
+  -d '{
+    "prompt": "Generate pipeline for PVeazie951/google-extention-ai-summarizer"
+  }'
+```
+
+### ✔ Expected Response
+```json
+{
+  "success": true,
+  "requires_confirmation": true,
+  "message": "A pipeline has been generated. Would you like me to commit this workflow file?",
+  "generated_yaml": "...",
+  "pipeline_metadata": { ... }
+}
+```
+
+The agent stores the repo name + generated YAML internally for the next step.
+
+---
+
+## 📌 Step 2 — Confirm the Commit
+Once the agent asks for confirmation, send a follow-up prompt such as:
+
+```bash
+curl -X POST http://localhost:3000/agent/wizard/ai \
+  -H "Content-Type: application/json" \
+  -H "Cookie: mcp_session=YOUR_SESSION_TOKEN" \
+  -d '{
+    "prompt": "yes commit"
+  }'
+```
+
+### ✔ Expected Response
+```json
+{
+  "success": true,
+  "tool_called": "pipeline_commit",
+  "committed_repo": "PVeazie951/google-extention-ai-summarizer",
+  "committed_path": ".github/workflows/ci.yml",
+  "sha": "...",
+  "html_url": "https://github.com/.../ci.yml"
+}
+```
+This confirms the workflow file was successfully committed to the repo.
+
+---
+
+## 🛡 How Confirmation Works
+To avoid accidental commits:
+
+- Only explicit intent triggers `pipeline_commit` (e.g., **"yes commit", "commit workflow", "apply pipeline"**)  
+- The agent now **ignores model-generated phrases** like "recent commits" or "show commits" during confirmation
+- User intent always overrides the agent's reasoning
+
+This makes the multi-step flow consistent and safe.
+
+---
+
+## 📌 Extra: Listing Repositories
+You can ask the agent anything:
+
+```bash
+curl -X POST http://localhost:3000/agent/wizard/ai \
+  -H "Content-Type: application/json" \
+  -H "Cookie: mcp_session=YOUR_SESSION_TOKEN" \
+  -d '{ "prompt": "List my repositories" }'
+```
+
+---
+
+## ✅ Summary
+- Use the Wizard Agent API to generate pipelines from natural language prompts
+- The agent **requires confirmation** before committing workflows
+- Multi-step flow works via persistent session state
+- Commits are safe and deterministic due to improved intent detection
+
+If you need a longer doc, diagrams, or frontend usage examples, I can generate those too.
+