github.com/suleymanozkeskin/collie

Under optimised search

ripgrep on a 29K-file repo (Kubernetes). Each query re-walks and re-searches candidate files.

Fresh run on the current build: rg still lands in the 400–880 ms range. Collie stayed under 14 ms on the lexical sample.
grep/rg also have no understanding of code structure. Just raw text.

What is Collie?

A lexical and structural search CLI that indexes your codebase once, keeps it current
while the daemon runs, and auto-stops when you're done.

Not a replacement for traditional search, specialized for agentic development, indexed per project / worktree.

How it works

# Start the daemon (indexes + watches for changes)
$ collie watch .

# Search instantly
$ collie search "handler"

# Stop when done (or auto-stops after 30min idle)
$ collie stop .

Index once → query faster → incremental updates via FSEvents

Four search modes

$ collie search "kube%"           # prefix
$ collie search "%config"         # suffix
$ collie search "handle request"  # multi-term AND

$ collie search "kind:fn handler"
$ collie search "kind:struct lang:go"
$ collie search "kind:method qname:Server::run"

$ collie search -e "func\s+\w+Handler"
$ collie search -e "TODO|FIXME|HACK"  # multi-pattern
$ collie search -e "impl.*Error" -i  # case-insensitive

$ collie search "kind:fn %Handler" --symbol-regex '\*.*Server'

Structural query vocabulary

kind:fn|method|class|struct|enum|interface|trait
     variable|field|property|constant|module|type|import
lang:go|rust|python|typescript|c|cpp|java|csharp|ruby|zig
path:src/api/              # scope to directory
qname:Server::run          # qualified name

$ collie search "kind:fn lang:go path:pkg/api/ init"
$ collie search "kind:struct Config"
$ collie search "kind:method qname:HollowKubelet::Run"
$ collie search "kind:trait path:src/ %Error%"

14 symbol kinds. 10 languages. Powered by Tree-sitter AST parsing.

Search beyond source code

include_pdfs = true

$ collie search "authentication flow"
  src/auth/handler.go:14        func authFlow() {
  docs/design-spec.pdf:3        Section 4.2: Authentication Flow
  rfcs/auth-v2.pdf:1            RFC: Revised authentication flow

Design specs, RFCs, API docs, research papers — indexed alongside code.
grep and ripgrep can't do this.

Benchmark — Kubernetes repo, 28,903 files

Lexical search: Collie vs ripgrep

Fresh run on the current build, 5 measured runs per query. Collie stayed tight; rg still paid full-scan cost on every query.

Benchmark — Kubernetes repo

Visualized: kubelet search

Benchmark — Kubernetes repo

Symbol search (no rg equivalent)

Find functions, structs, methods — not just text matches.
Powered by Tree-sitter across 11 languages.

Benchmark — Kubernetes repo

Regex search (bounded top-k, index-accelerated)

Collie extracts literal fragments from your regex,
narrows candidate files via the index, then applies the full regex.

Honest framing: Collie regex is optimized for interactive bounded queries.
For exhaustive -n 0 scans, ripgrep still wins today.

Why this matters for agents

Agents know what they're looking for

$ rg "func \(.*Authorization.*\) Validate"          522 ms

$ collie search "kind:method path:pkg/kubeapiserver/options/ %validate%"  72 ms

Agents already know the kind, language, and approximate location.
Structural queries turn that knowledge into precise results.

New capability

Symbol + Regex: --symbol-regex

Narrow with structure first, refine with regex. No rg equivalent.

$ collie search "kind:fn %Handler" --symbol-regex '\*.*Server'
$ collie search "kind:method qname:Server::" --symbol-regex 'Handler'
$ collie search "kind:fn %validate%" --symbol-regex 'webhook|Webhook'

The more structural info in the symbol query, the faster the regex refinement.

Agentic benchmark — per-task

Task breakdown (time to first hit)

Built for AI agents

$ collie search "kind:fn handler" \
    --format json -n 5
{
  "type": "symbol",
  "count": 5,
  "results": [{
    "path": "pkg/api/handler.go",
    "line": 42,
    "kind": "function"
  }, ...]
}
$ echo $?
0

Under the hood

All Rust. Single binary. No runtime dependencies.

11 languages

Go • Rust • Python • TypeScript • JavaScript
C • C++ • Java • C# • Ruby • Zig

Symbol extraction: functions, methods, structs, classes, enums,
interfaces, traits, variables, fields, constants, modules, type aliases, imports

Indexing performance

One-time cost. After that, notify keeps the index current incrementally.

Get started

$ cargo install collie-search

$ collie watch .              # start daemon, index repo
$ collie status .             # verify it's running

$ collie search "handleRequest"
$ collie search "kind:fn path:src/ init"
$ collie search -e "func\s+\w+Handler"
$ collie search "kind:fn %Handler" --symbol-regex '\*.*Server'
$ collie search "config" --format json  # for agents

$ collie skill               # prints reference card for LLM context

github.com/suleymanozkeskin/collie