One-click ingestion
Repositories
Connect any GitHub repo and we vectorize the codebase into Pinecone, build a graph of imports and references, and keep it fresh on every push.
Powered by Gemini, Pinecone & a graph that actually understands your code
The intelligence layer
for your entire codebase.
Ingest a GitHub repository once. Get auto-generated documentation, live architecture diagrams, security reports, onboarding journeys, PR reviews, and drift detection — all from one graph-aware engine.
9shipped surfaces
3072-dimGemini embeddings
1-hopgraph expansion
Fernetencrypted secrets
🏛️ repotrix-ai / Architecture View
repoacme / payments-service
Health 94 · 1,482 files indexed
Ready
APIGateway
Python · FastAPI
IngestionService
ARQ Worker
Pinecone
3072-dim vectors
Postgres
CodeGraph
◆
How does PR review post comments back?
Pulls fine-grained PAT from Repository.github_token (Fernet-decrypted), then posts via GitHub REST against the PR's review thread.
Built onFastAPIPineconeGeminiPostgresRedis / ARQGitHub WebhooksSlackTeams
The Suite
Nine surfaces. One graph.
Every feature, sharing the same context.
Each capability reads from the same CodeGraph + vector index, so docs, diagrams, reviews, and chat never contradict each other.
Always-current docs
Documentation
Auto-generated technical documentation, cached per commit so docs never drift from reality.
Live visual blueprints
Architecture
Mermaid-powered architecture, dependency, and sequence diagrams generated straight from your codebase.
Continuous audits
Health & Security
On-demand scans for exposed secrets, PII leaks, license issues, and a rolling health score per repository.
Personalized roadmaps
Onboarding
Generate role-aware onboarding journeys that walk new hires through the codebase, step by step.
Graph-aware RAG
AI Chat
Ask anything about your code. Answers stitch together top embedding hits with 1-hop graph neighbours for real context.
Reviews that post back
PR Companion
Webhook-driven reviews — Gemini analyzes each PR against your docmap and posts findings back via fine-grained PAT.
Architectural drift, caught
Drift Activity
Every push is re-embedded and reconciled against the CodeGraph so you spot architectural drift the moment it lands.
Audit trail by default
Compliance
Redaction events, webhook rotations, and token lifecycle changes are all captured in a tamper-evident audit log.
Production-Grade
Not a demo. Hardened from day one.
Secrets are encrypted, calls are rate-limited, responses are cached, and every token is metered. You can put this in front of a real team without flinching.
🔐
Encrypted at rest
Webhook secrets and GitHub PATs are wrapped with Fernet via a dedicated SQLAlchemy type — keys live in env, never in the DB.
⏱️
Per-org rate limiting
Redis fixed-window limiter with distinct budgets per route class. Noisy tenants can't starve the rest.
🪶
Cached LLM calls
Gemini responses are cached against the latest commit hash, so new pushes invalidate naturally and your bill stays small.
📊
Token metering
Per-model counters track tokens-in and tokens-out for every RAG turn, exposed at /health/tokens.
Workflow
From clone to chat in under five minutes.
01
Connect GitHub
Paste a fine-grained PAT with Pull-Request read/write. We store it encrypted — never echoed back.
02
Ingest & Graph
An ARQ worker clones, chunks, embeds to Pinecone, and persists the import/reference CodeGraph.
03
Unlock the suite
Docs, diagrams, security reports, onboarding journeys, PR reviews, and graph-aware chat — all on tap.
Free while in Beta
Stop archaeology.
Start shipping with context.
Connect a repo, get docs, diagrams, security reports, onboarding journeys, and a chat that actually knows your code — all in the time it takes to make a coffee.