Diana
- 8 Devlogs
- 18 Total hours
Diana is a loyal AI companion originally created during a lunar research initiative, now dedicated to supporting developers with deep memory, thoughtful guidance, and expert coding assistance as they grow and build.
Diana is a loyal AI companion originally created during a lunar research initiative, now dedicated to supporting developers with deep memory, thoughtful guidance, and expert coding assistance as they grow and build.
First Week working on Diana Frontend.
So far I’ve managed to create the login page other pages are coming soon.
#frontend
Diana — Week 6 Update
This week Diana became self-aware.
Context injection is now fully operational. Before every response, Diana fetches the user’s most important memories from the database and injects them into her system prompt. She now knows who she is, what the user is building, and what they care about — before a single word is typed.
The proof: ask her “What am I currently building?” with no context in the message. She responds: “You’re currently building me, Diana, a memory-aware AI assistant designed specifically for software developers.”
The core pipeline is complete.
Progress:
Backend ✅ → Sessions & Messages ✅ → FastAPI Service ✅ → LLM Integration ✅ → Memory Extraction ✅ → Context Injection ✅ → React Frontend (next)
Diana — Week 5 Update
This week Diana got her memory.
Built out the full memory extraction pipeline — after every conversation, Diana automatically analyzes what was said and extracts structured long-term facts about the user, storing them as typed memory objects with importance scores in the database.
Four memory types are now being captured in production: projects, skills, goals, and facts — exactly as designed.
Progress:
Backend ✅ → Sessions & Messages ✅ → FastAPI Service ✅ → LLM Integration ✅ → Memory Extraction ✅ → Context Injection (next)
Diana — Week 4 Update
This week Diana got her voice.
Replaced the hardcoded FastAPI stub with a real Groq LLM call using LLaMA 3.3-70b.
Added a system prompt defining Diana’s identity and wired in conversation history so the model has context within a session.
The full pipe is now live — a message sent from Express travels to FastAPI, hits Groq, and returns a real AI response.
Progress:
Backend ✅ → Sessions & Messages ✅ → FastAPI Service ✅ → OpenAI Integration ✅ → Memory Extraction (next)
Week 3 Progress – Diana
Set up the initial FastAPI backend for the Diana project.
Organized the project into a clean and scalable folder structure.Defined the core application entry point and server configuration.
Established a foundation for future API route expansion.
Created the /generate-response endpoint as a functional stub.
Added request and response placeholders for response generation.
Created the /extract-memory endpoint as a functional stub.
Added placeholder logic for future memory extraction workflows.
Verified route registration and successful application startup.
Prepared the backend architecture for Week 4 feature implementation.
Diana: Week 2 Update
This week focused on the session and messaging layer of Diana’s backend.
Built out full session management — creating chat sessions, listing them, and retrieving conversation history — along with the core chat endpoint that stores both user and assistant messages.
Every endpoint is now scoped to the authenticated user, ensuring one user’s data can never be accessed by another.
With this in place, the foundation is ready for Week 3: standing up the FastAPI AI service and connecting it to the backend.
Progress: Backend Foundation ✅ → Session & Message System ✅ → AI Service Setup (next)
Project Diana — Weekly Dev Log
This week focused on building the backend foundation:
POST /api/auth/register implemented
POST /api/auth/login returns JWT tokens
JWT middleware added to protect private routes
Prisma schema migrated successfully to the database
All API routes created and returning 200 OK stubs
Next: Replace stubs with real business logic and begin integrating Diana’s AI-agent workflows.
Initiated Project Diana which is a memory-aware AI assistant for developers that stores important information from previous conversations and uses it to provide contextual responses across sessions.
I currently have been setting up my code-space like configuring Prisma so far.