sampler-cli 0.4.3

Token-efficient CLI for indexing and searching code symbols (Python-first, designed for minimal LLM/agent context size)

Sampler

Token-efficient CLI for indexing and searching code symbols across multiple projects.

Current version: 0.4.3

Designed for humans and agents: compact default output, short paths, and low-noise symbol views.

Requirements

• Python 3.11+

Installation

pip install sampler-cli

Development setup:

pip install -e '.[dev]'

Semantic stack (TF-IDF + local hash fallback):

pip install -e '.[semantic]'

Quick Start

sampler init
sampler project add myproj /absolute/path/to/project --language auto
sampler index myproj
sampler search retry --project myproj
sampler symbols myproj
sampler overview src/main.py

Command Overview

Core:

• sampler version [--plain]
• sampler init
• sampler index <project>
• sampler search <query> [--project <name>] [--type <t>] [--limit <n>] [--semantic] [--style plain|bars]
• sampler search-all <query> [--type <t>] [--limit <n>]
• sampler symbols <project> [--type <t>] [--limit <n>]
• sampler overview <filepath> [--style plain|bars]

Relationships:

• sampler callers <symbol> [--project <name>] [--file <path-or-suffix>]
• sampler usages <symbol> [--project <name>] [--file <path-or-suffix>]
• sampler related <symbol> [--project <name>] [--file <path-or-suffix>] [--style plain|bars]
• Selector alternativo: <path>:<symbol> (ej. app/utils/helpers.py:format_kda)

Project management:

• sampler project add <name> <path> --language <python|go|typescript|javascript|auto>
• sampler project update <name> [--path <abs-path>] [--language <lang>]
• sampler project list
• sampler project deps <name>
• sampler project remove <name>

Config:

• sampler config show
• sampler config embeddings [--provider P] [--model M]

Semantic and analysis:

• sampler embed <project> [--batch-size <n>]
• sampler stale-code <project> [--limit <n>]

Embeddings & Semantic Search

sampler search --semantic (and hybrid ranking) supports pluggable providers via the adapter pattern:

• Default: bge-small (BAAI/bge-small-en-v1.5 via fastembed — lightweight ONNX, ~384 dim, local).
• Other built-ins: hash (always-on deterministic fallback), ollama (e.g. nomic-embed-text), nomic, openai, fastembed.
• TF-IDF (sklearn, on-the-fly, no pre-embed) remains the fast lexical primary when no provider embeddings are precomputed for the active model.
• Hash fingerprint is the final always-available fallback.

Configuration (in ~/.sampler/config.yaml or via sampler config embeddings ...):

embeddings:
  provider: "bge-small"
  # provider: "ollama"
  # model: "nomic-embed-text"
  # base_url: "http://localhost:11434"

Install:

# For default BGE (recommended for most users)
pip install 'sampler-cli[embeddings]'

# Or for Ollama / OpenAI only
pip install 'sampler-cli[ollama-embeddings]'
pip install 'sampler-cli[openai-embeddings]'

sampler embed <project> precomputes vectors using the current configured provider (progress bar). Changing provider? Re-run embed after updating config (old vectors are ignored until re-embedded).

Offline / air-gapped: provider: hash (or just don't install the embeddings extra — TF-IDF + hash still work if you have [semantic]).

Language Support

• Python parser: stdlib AST (stable)
• Go parser: tree-sitter-go (real extraction)
• TypeScript/JavaScript parser: tree-sitter-typescript (real extraction)
• --language auto: per-file language detection for monorepos/multi-language projects

Stale Code Detection

sampler stale-code <project> reports candidate stale functions/methods where:

• function is called from test files
• function has zero non-test callers in project call graph
• symbol is defined in production code (symbols defined in test files are excluded)

Test file detection supports common multi-language patterns:

• Python: tests/, test_*.py, *_test.py
• Go: *_test.go
• TypeScript/JavaScript: __tests__/, test/, spec/, *.test.*, *.spec.*

This is heuristic signal, not guaranteed dead-code proof.

Examples

$ sampler search worker --project myproj
myproj:src/tasks/celery_app.py:70 function on_worker_ready  def on_worker_ready(sender)

$ sampler related ConfigManager --project myproj --style bars
myproj:src/config.py:24-105 class ConfigManager  [parent]
...

$ sampler stale-code myproj
myproj:src/utils/retry.py:12-28 function retry_request  test_callers=2 non_test_callers=0  [tests.test_retry.test_retry_request]

Data Location

• Config: ~/.sampler/config.yaml
• DB: ~/.sampler/graph.db

Running Tests

pytest -q

Notes

• Compact output is default by design (token-efficient for agent workflows).
• For broader roadmap details, see TODO.md and PLAN.md.