What I Built
Codebase Guide is a conversational AI assistant that helps new developers understand and safely navigate complex multi-repository codebases. Instead of spending hours hunting through repos and asking seniors "where do I start?", juniors can ask natural language questions like "Where is authentication handled?" or "How do I add a new profile field?" and get instant, structured answers with files, repos, and test commands.
The problem: onboarding onto large, multi-service systems is painful. Documentation is scattered, tribal knowledge lives in senior devs' heads, and juniors waste days just figuring out where to add code.
Codebase Guide solves this by indexing services, patterns, and playbooks across all repos, then using Algolia Agent Studio to retrieve the right context and generate mentor-style guidance.
Demo
Live UI:
https://codebase-guide-final.vercel.app
video:
https://youtu.be/RlgZvAfyikU?si=E9raVWfmRduM8DY-
GitHub:
https://github.com/pulipatikeerthana9-wq/codebase-guide-final
How I Used Algolia Agent Studio
I created three specialized indices to power fast, contextual retrieval:
services_index: Maps each service/repo to its purpose, tech stack, owner team, entry files, and key directories. Tags like auth, payments, frontend enable quick filtering.
patterns_index: Stores "how we do X" patterns—authentication middleware, error handling, feature flags, webhook processing—with code snippets and explanations.
playbooks_index`: Step-by-step guides for common tasks: "Add a new profile field," "Create a protected route," "Add a notification type." Each includes repos involved, exact steps, and test commands.
The Agent Studio configuration:
System prompt: Positioned the agent as a "senior dev mentor" who always answers in 4 parts: current implementation, files to inspect, safe change plan, tests to run.
**Retrieval tools: **Configured Algolia Search across all three indices with tag-based filtering (auth, payments, profiles, etc.).
Structured output: The agent retrieves relevant services, patterns, and playbooks, then synthesizes them into actionable guidance.
Example query: "How do I add a new profile field in API and frontend?"
→ Agent retrieves:
users-service from services_index
frontend-app from services_index
pb_add_profile_field playbook from playbooks_index
→ Returns: files to touch, database migration steps, validation updates, and test commands.
Why Fast Retrieval Matters
Without fast, structured retrieval, juniors either:
Grep through hundreds of files (slow, overwhelming)
Interrupt seniors constantly (blocks their work)
Make unsafe changes because they didn't find the right pattern
With Algolia's sub-second retrieval across three indices:
Questions that took 30+ minutes to answer now take 10 seconds.
Juniors get complete context (services + patterns + playbooks) in one response.
Try It Yourself
The agent can filter by tags (auth, backend, frontend) to surface exactly what's needed, not every file that mentions "user."
This turns onboarding from a week-long slog into a guided, self-serve experience.
The agent is currently in draft mode in Algolia Agent Studio. To use it live with your own queries:
Fork the GitHub repo
Clone the Algolia indices (or create your own with your codebase data)
In Agent Studio, create a provider profile with your own LLM API key (OpenAI, Anthropic, or Gemini)
Publish the agent and embed it in the UI
The UI is deployed at https://codebase-guide-final.vercel.app and shows the complete interface design. The retrieval logic and agent configuration are fully functional and can be tested in the Algolia playground.
Top comments (0)