How Documentation Search Works

From code to searchable documentation in seconds

From Code to Searchable Documentation

Automated pipeline powered by Aider AI and Typesense

1
Repository
Your codebase
Connect your Git repo
2
Aider AI
Code analysis
Generate documentation
3
Markdown
Structured docs
Framework-specific
4
Parser
H2 sections
Split by headings
5
Typesense
Search index
User-scoped collections
6
MCP Search
AI queries
<50ms results
Repositories
<50ms
Search Time
25x
More Efficient
Auto
Reindexing

25x More Token-Efficient

Stop wasting 98% of your context on irrelevant content

Traditional Approach

Load entire documentation

Tokens Loaded
50,000
Entire documentation file
Cost per Query
$0.50
Based on input tokens
Wastes 98% of context on irrelevant content
Expensive token usage on every request
Slow processing of large context windows
Hits context limits quickly

ULPI Approach

Semantic search

Tokens Loaded
2,000
5 relevant sections only
Cost per Query
$0.02
25x cheaper than traditional
Only loads relevant sections (98% savings)
25x more cost-effective per query
Sub-50ms search latency
Never hits context limits
25x

Semantic Search

Typesense-powered search returns relevant documentation sections in under 50ms. Vector similarity and full-text search combined for best results.

25x Token Efficiency

Load only 2,000 tokens of relevant sections instead of 50,000 tokens of full docs. 98% reduction in context usage and costs.

Auto-Reindexing

Webhooks keep documentation current. Git push triggers delta updates in 1.2 seconds. Full reindex for major changes.

Multi-Repository

Index unlimited repositories and branches. User-scoped Typesense collections for privacy. Framework-agnostic indexing.

Detail Levels

Choose snippet (30% content, 20 results), standard (full, 10 results), or comprehensive (full, 30 results) based on your needs.

AI-Generated Docs

Aider AI generates documentation from your codebase using framework-specific templates. Automatic code-to-docs pipeline.

Search Results Preview

Example search results showing relevance scoring and token efficiency

Try:
Found 3 results in 48ms
Total tokens: 1015

OAuth 2.0 Implementation

backend-api
95%
420 tokens

Complete OAuth 2.0 authentication flow with JWT tokens, refresh tokens, and secure session management...

API Authentication Middleware

backend-api
89%
315 tokens

Middleware for validating bearer tokens, handling expired sessions, and enforcing role-based access control...

Social Login Integration

frontend-web
82%
280 tokens

Integration with Google, GitHub, and Microsoft identity providers using OAuth 2.0 authorization code flow...

Auto-Reindexing via Webhooks

Documentation stays current automatically. Delta updates complete in 1.2 seconds.

0ms
Git Push
Developer commits code
12ms
Webhook Received
GitHub sends event
850ms
Delta Update
Only changed files
1.2s
Reindexed
Documentation updated
Delta Update
3 files changed
src/auth/oauth.ts
modified+42 -8
docs/authentication.md
modified+15 -3
src/middleware/auth.ts
added+67
Full Reindex
Time:~8.5s
Files:All files
When:Initial setup
Delta Update
Time:~1.2s
Files:Changed only
When:Every push
7x faster
Average reindex time: 1.2 seconds

Real-World Use Cases

See how teams use documentation search to move faster

API Documentation Search

Find authentication examples, error handling patterns, or rate limiting configs instantly across all your documentation.

"Show me OAuth 2.0 implementation" → 3 relevant sections in 48ms

Onboard New Developers

New team members get instant answers from your codebase documentation. No more digging through wikis or asking teammates.

"How does our payment processing work?" → Complete flow documentation

Multi-Agent Knowledge Sharing

Multiple AI agents can search the same documentation without duplicating context. All agents cite the same source of truth.

Claude, Copilot, and Cursor all reference the same auth docs

Code Review Automation

AI code review bots can cite relevant documentation sections when suggesting improvements or catching violations.

"This violates our error handling pattern (see docs section 4.2)"

ChatOps Integration

Slack or Discord bots answer technical questions by searching your documentation. Sub-50ms responses feel instant.

"/docs deployment procedure" → Step-by-step guide in Slack

PR Descriptions

Auto-generate pull request descriptions with relevant documentation sections. Help reviewers understand context quickly.

PR touches auth → Auto-includes OAuth flow documentation

Framework Agnostic

Works with Laravel, Next.js, React, Vue, Django, and any other framework

Framework Agnostic

Works with any codebase. Aider AI adapts to your framework automatically.

Laravel
1,234 docs
Next.js
892 docs
React
756 docs
Vue.js
654 docs
Django
543 docs
Ruby on Rails
487 docs
Express.js
412 docs
NestJS
389 docs
FastAPI
321 docs
Spring Boot
298 docs
ASP.NET
267 docs
Flask
234 docs
6,487
Total documentation sections indexed

3 MCP Tools

Simple API for semantic documentation search

search_documentation

Semantic search with repository/branch filters and detail levels

search_documentation({
  query: "authentication examples",
  detail_level: "standard",
  repository_id: "backend-api",
  limit: 10
})

expand_document

Get full content of a specific documentation section by ID

expand_document({
  document_id: "doc_abc123"
})

list_repositories

List all indexed repositories with document counts

list_repositories()
// Returns: [{id: 1, name: "backend-api", docs: 234}, ...]

Ready to Give Your Agents Instant Codebase Knowledge?

Stop loading entire documentation into context. Let AI agents search semantically and find exactly what they need. Join teams shipping 25x more efficiently.