Skip to content

Commit af0d9bb

Browse files
prosdevclaude
prosdev
and
claude
committed
docs(website): audit and fix all stale claims on doc site
- Rewrite homepage: drop unverified benchmark claims (42% savings, 99% fewer tokens), drop "context bundling" framing (dev_plan removed), replace with factual descriptions of what tools actually do - Fix LanceDB → Antfly in dev-search, dev-map, dev-health, dev-status tool pages - Fix all-MiniLM-L6-v2 → BAAI/bge-small-en-v1.5 in docs/index and dev-search - Fix dev explore → dev search --similar-to in cli.mdx - Add Antfly install to quick start (was missing) - Fix test count to 1,600+ Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
1 parent 3b9a6d8 commit af0d9bb

7 files changed

Lines changed: 51 additions & 115 deletions

File tree

‎website/content/docs/cli.mdx‎

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -168,8 +168,8 @@ dev index
168168
# Search for code
169169
dev search "user authentication"
170170

171-
# Explore code patterns
172-
dev explore pattern "error handling"
171+
# Find similar code
172+
dev search --similar-to src/utils/helpers.ts
173173
```
174174

175175
### Integration with Cursor

‎website/content/docs/index.mdx‎

Lines changed: 3 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -1,12 +1,12 @@
11
# Introduction
22

3-
**dev-agent** provides semantic code search and context bundling to AI assistants like Cursor and Claude Code via MCP.
3+
**dev-agent** provides semantic code search to AI assistants like Cursor and Claude Code via MCP.
44

5-
We built this for ourselves. When exploring large codebases, we found AI tools spending too much time grepping through files and reading entire files to find relevant code. dev-agent gives them a faster path: search by meaning, get code snippets, bundle context.
5+
We built this for ourselves. When exploring large codebases, we found AI tools spending too much time grepping through files and reading entire files to find relevant code. dev-agent gives them a faster path: search by meaning and get code snippets.
66

77
## What it does
88

9-
1. **Indexes your codebase** locally with embeddings (all-MiniLM-L6-v2)
9+
1. **Indexes your codebase** locally with embeddings (BAAI/bge-small-en-v1.5)
1010
2. **Returns code snippets** — not just file paths, reducing input tokens by 99%
1111
3. **Provides semantic search** — find code by meaning, not just keywords
1212

‎website/content/docs/tools/dev-health.mdx‎

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -48,7 +48,7 @@ Components:
4848
┌─────────────────┬────────┬──────────┬─────────────────────┐
4949
│ Component │ Status │ Latency │ Details │
5050
├─────────────────┼────────┼──────────┼─────────────────────┤
51-
│ Vector Storage │ ✅ Pass │ 24ms │ LanceDB connected │
51+
│ Vector Storage │ ✅ Pass │ 24ms │ Antfly connected │
5252
│ Repository Index│ ✅ Pass │ 12ms │ 1,832 components │
5353
│ GitHub Index │ ✅ Pass │ 8ms │ 172 documents │
5454
│ Rate Limiter │ ✅ Pass │ <1ms │ 89/100 tokens │
@@ -70,7 +70,7 @@ CPU: 2.3%
7070

7171
| Component | What It Checks |
7272
|-----------|----------------|
73-
| **Vector Storage** | LanceDB connection and query capability |
73+
| **Vector Storage** | Antfly connection and query capability |
7474
| **Repository Index** | Index exists and is readable |
7575
| **GitHub Index** | GitHub metadata is accessible |
7676
| **Rate Limiter** | Token bucket has capacity |

‎website/content/docs/tools/dev-map.mdx‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -34,7 +34,7 @@ This helps AI assistants quickly understand codebase organization and activity.
3434

3535
## Hot Paths (most referenced)
3636
1. `packages/core/src/indexer/index.ts` (RepositoryIndexer) - 47 refs
37-
2. `packages/core/src/vector/store.ts` (LanceDBVectorStore) - 32 refs
37+
2. `packages/core/src/vector/store.ts` (AntflyVectorStore) - 32 refs
3838
3. `packages/mcp-server/src/server/mcp-server.ts` (MCPServer) - 28 refs
3939

4040
## Directory Structure

‎website/content/docs/tools/dev-search.mdx‎

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -77,8 +77,8 @@ This surfaces test files without affecting semantic search rankings — tests ar
7777

7878
## How It Works
7979

80-
1. **Query embedding** — Your query is converted to a vector using MiniLM-L6-v2
81-
2. **Vector search** — LanceDB finds semantically similar code components
80+
1. **Query embedding** — Your query is converted to a vector using BAAI/bge-small-en-v1.5
81+
2. **Vector search** — Antfly finds semantically similar code components
8282
3. **Ranking** — Results are sorted by cosine similarity score
8383
4. **Formatting** — Results are formatted for LLM context windows
8484

‎website/content/docs/tools/dev-status.mdx‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -52,7 +52,7 @@ Index Duration: 12.3s
5252
```
5353
## Index Status
5454
55-
Vector Storage: LanceDB
55+
Vector Storage: Antfly
5656
Location: ~/.dev-agent/indexes/my-project
5757
Size: 24.5 MB
5858
Components: 1,832

‎website/content/index.mdx‎

Lines changed: 40 additions & 104 deletions
Original file line numberDiff line numberDiff line change
@@ -2,7 +2,7 @@
22
title: dev-agent
33
---
44

5-
import { Callout, Steps, Tabs, FileTree } from 'nextra/components'
5+
import { Callout, Steps, Tabs } from 'nextra/components'
66
import { latestVersion } from './latest-version'
77

88
# dev-agent
@@ -11,87 +11,26 @@ Local semantic code search for Cursor and Claude Code via MCP.
1111

1212
[Get Started](/docs) · [View on GitHub](https://github.com/prosdevlab/dev-agent)
1313

14-
<Callout type="warning">
15-
**Project moved.** dev-agent started as a hack project at [Lytics](https://github.com/lytics/dev-agent) and is now maintained independently at [prosdevlab/dev-agent](https://github.com/prosdevlab/dev-agent). All future development happens here.
16-
</Callout>
17-
1814
<Callout type="info">
1915
**New in v{latestVersion.version}** — {latestVersion.summary} [See what's new →](/updates)
2016
</Callout>
2117

2218
<Callout type="default">
23-
**Built by engineers, for engineers.** An MCP server that gives your AI tools semantic code search and context bundling. Savings scale with task complexity — up to 42% on debugging tasks.
19+
**Built by engineers, for engineers.** An MCP server that gives your AI tools semantic code search — find code by meaning, not keywords. Everything runs locally. Your code never leaves your machine.
2420
</Callout>
2521

26-
## Why it saves money
27-
28-
dev-agent doesn't just search — it **bundles context** so Claude reads less:
29-
30-
| What dev-agent does | Manual equivalent | Savings |
31-
|---------------------|-------------------|---------|
32-
| Returns code snippets in search | Read entire files | 99% fewer input tokens |
33-
| Semantic search finds relevant code | grep chains + manual filtering | 42% cost reduction |
22+
## What it does
3423

35-
**The harder the task, the bigger the savings.**
36-
37-
## Measured results by task type
38-
39-
| Task Type | Cost Savings | Time Savings | Why |
40-
|-----------|--------------|--------------|-----|
41-
| **Debugging** | **42%** | 37% | Semantic search beats grep for "where is the bug?" |
42-
| **Exploration** | **44%** | 19% | Find code by meaning, not keywords |
43-
| **Implementation** | **29%** | 22% | Semantic search finds relevant code |
44-
| **Simple lookup** | ~0% | ~0% | Both approaches are fast |
45-
46-
<Callout type="warning">
47-
**Trade-off:** Faster but sometimes less thorough. Baseline Claude provides more diagnostic shell commands. dev-agent excels when you need to explore or understand code.
48-
</Callout>
24+
Your AI tool gets 6 MCP tools that understand your codebase:
4925

50-
## Same question, different approach
51-
52-
We asked Claude Code: *"Debug why search returns duplicates"*
53-
54-
<Tabs items={['Without dev-agent', 'With dev-agent']}>
55-
<Tabs.Tab>
56-
**Claude's approach:**
57-
<FileTree>
58-
<FileTree.Folder name="grep 'duplicate' → 30 matches" />
59-
<FileTree.Folder name="grep 'search' → 100+ matches" />
60-
<FileTree.Folder name="grep 'id' → too many, narrow down" />
61-
<FileTree.File name="Read indexer/index.ts (441 lines)" />
62-
<FileTree.File name="Read vector/store.ts (258 lines)" />
63-
<FileTree.File name="Read scanner/typescript.ts (full file)" />
64-
<FileTree.File name="Read scanner/markdown.ts (full file)" />
65-
<FileTree.Folder name="... more greps and reads" />
66-
</FileTree>
67-
68-
**Result:** 18+ tool calls, 10 files read → **$1.37, 12 minutes**
69-
70-
*18,800 input tokens consumed*
71-
</Tabs.Tab>
72-
<Tabs.Tab>
73-
**Claude's approach:**
74-
<FileTree>
75-
<FileTree.Folder name="dev_search 'search duplicate results'" defaultOpen>
76-
<FileTree.File name="→ store.ts (with upsert code snippet)" />
77-
<FileTree.File name="→ indexer.ts (with ID generation)" />
78-
</FileTree.Folder>
79-
<FileTree.Folder name="dev_search 'document ID generation'" defaultOpen>
80-
<FileTree.File name="→ typescript.ts (ID pattern)" />
81-
<FileTree.File name="→ markdown.ts (slug generation)" />
82-
</FileTree.Folder>
83-
<FileTree.File name="Read store.ts (for detail)" />
84-
</FileTree>
85-
86-
**Result:** 6 tool calls, 5 files read → **$0.79, 7.5 minutes**
87-
88-
*65 input tokens consumed (99.7% less)*
89-
</Tabs.Tab>
90-
</Tabs>
91-
92-
<Callout type="info">
93-
**Same root cause identified. 42% cheaper. 37% faster.**
94-
</Callout>
26+
| Tool | What it does |
27+
|------|--------------|
28+
| [`dev_search`](/docs/tools/dev-search) | Hybrid search (BM25 + vector + RRF) — returns code snippets, not just file paths |
29+
| [`dev_refs`](/docs/tools/dev-refs) | Find callers/callees of any function |
30+
| [`dev_map`](/docs/tools/dev-map) | Codebase structure with hot paths (most referenced files) |
31+
| [`dev_patterns`](/docs/tools/dev-patterns) | Compare coding patterns against similar files |
32+
| [`dev_status`](/docs/tools/dev-status) | Repository indexing status and health |
33+
| [`dev_health`](/docs/tools/dev-health) | Server health checks |
9534

9635
## How it works
9736

@@ -117,21 +56,37 @@ flowchart LR
11756
D <--> E
11857
```
11958

120-
**Key insight:** dev-agent returns **code snippets with context** using hybrid search (keyword matching + semantic understanding) — Claude doesn't read entire files. This is why input tokens drop by 99%.
59+
1. **Index** — Scanner parses your code (ts-morph for TypeScript, tree-sitter for Go), extracts functions, classes, types, imports, and call graphs
60+
2. **Embed** — Antfly generates vector embeddings locally via Termite (ONNX, BAAI/bge-small-en-v1.5)
61+
3. **Search** — Hybrid search combines BM25 keyword matching with vector similarity, fused via Reciprocal Rank Fusion (RRF)
62+
4. **Auto-update** — File watcher detects changes and re-indexes automatically while the MCP server is running
63+
64+
## Why it helps
65+
66+
Instead of grep chains and reading entire files, your AI tool can ask conceptual questions:
67+
68+
| Question | Without dev-agent | With dev-agent |
69+
|----------|-------------------|----------------|
70+
| "Where do we handle auth?" | `grep -r "auth"` → 200 matches | `dev_search "authentication"` → 5 ranked results with code |
71+
| "What calls this function?" | Manual file reading | `dev_refs "validateUser"` → callers + callees |
72+
| "What's in this codebase?" | `ls`, `find`, read READMEs | `dev_map` → structure with component counts |
73+
74+
The key difference: semantic search finds code by **meaning**, not text matching. "Where do we handle errors?" finds try/catch blocks, error middleware, and validation functions — even if none contain the word "error."
12175

12276
## Quick Start
12377

12478
<Steps>
12579
### Install
12680

12781
```bash
128-
npm install -g dev-agent
82+
npm install -g @prosdevlab/dev-agent
83+
brew install --cask antflydb/antfly/antfly
12984
```
13085

131-
### Index your repository
86+
### Setup and index
13287

13388
```bash
134-
cd your-project
89+
dev setup
13590
dev index
13691
```
13792

@@ -143,37 +98,18 @@ dev mcp install # For Claude Code
14398
```
14499
</Steps>
145100

146-
## 6 MCP Tools
147-
148-
| Tool | What it does |
149-
|------|--------------|
150-
| [`dev_search`](/docs/tools/dev-search) | Semantic code search — returns snippets, not just paths |
151-
| [`dev_refs`](/docs/tools/dev-refs) | Find callers/callees of any function |
152-
| [`dev_map`](/docs/tools/dev-map) | Codebase structure with change frequency |
153-
| [`dev_patterns`](/docs/tools/dev-inspect) | Inspect files (compare implementations, check patterns) |
154-
| [`dev_status`](/docs/tools/dev-status) | Repository indexing status |
155-
| [`dev_health`](/docs/tools/dev-health) | Server health checks |
156-
157-
## When to use it
158-
159-
| Scenario | dev-agent? | Expected Savings |
160-
|----------|------------|------------------|
161-
| Debugging unfamiliar code | ✅ Yes | **42% cost** |
162-
| Exploring large codebase | ✅ Yes | **44% cost** |
163-
| Implementing GitHub issues | ✅ Yes | **29% cost** |
164-
| Small codebase you know | ❌ Skip | ~0% |
165-
| Need exhaustive file reads | ⚠️ Maybe | Trade speed for thoroughness |
166-
167101
## Features
168102

169-
- **Code Snippets** — Search returns code, not just file paths
103+
- **Hybrid Search** — BM25 keyword + vector semantic, fused with RRF
104+
- **Code Snippets** — Search returns actual code, not just file paths
105+
- **Call Graph** — Callers/callees extracted from AST at index time
170106
- **Multi-Language** — TypeScript, JavaScript, Go, Markdown
171-
- **100% Local** — Your code never leaves your machine
172-
- **Sub-second Search** — Fast even on large repos with LanceDB
173-
- **1500+ Tests** — Production-grade reliability
107+
- **100% Local** — Antfly runs on your machine. No data leaves.
108+
- **Auto-Index** — File watcher re-indexes on save while MCP server runs
109+
- **1,600+ Tests** — Production-grade reliability
174110

175111
---
176112

177-
[Read the story behind dev-agent →](https://www.devbypros.com/blog/10-days-of-vibe-coding)
113+
MIT License
178114

179-
MIT License • Built by [prosdev](https://github.com/prosdev)
115+
[Read the story behind dev-agent →](https://www.devbypros.com/blog/10-days-of-vibe-coding)

0 commit comments

Comments
 (0)