### Install Semble via pip or uv

Source: https://github.com/minishlab/semble/blob/main/README.md

Install Semble using either pip or uv. This command makes the `semble` executable available in your environment.

```bash
pip install semble
```

```bash
uv tool install semble
```

--------------------------------

### Install Semble if not on PATH

Source: https://github.com/minishlab/semble/blob/main/src/semble/agents/semble-search.md

If the `semble` command is not found in your system's PATH, you can install and run it using `uvx`. This command ensures the `semble` tool with the `mcp` extra is available.

```bash
uvx --from "semble[mcp]" semble
```

--------------------------------

### Run Probe Benchmark

Source: https://github.com/minishlab/semble/blob/main/benchmarks/README.md

Benchmark using the 'probe' tool. Ensure 'probe' is installed globally (`npm install -g @buger/probe`). Allows specifying specific repositories.

```bash
uv run python -m benchmarks.baselines.probe
```

```bash
uv run python -m benchmarks.baselines.probe --repo fastapi --repo axios
```

--------------------------------

### Install Semble for Claude Code

Source: https://github.com/minishlab/semble/blob/main/README.md

Add Semble to Claude Code using the `claude mcp add` command. Ensure you have `uv` installed for managing dependencies.

```bash
claude mcp add semble -s user -- uvx --from "semble[mcp]" semble
```

--------------------------------

### Generate Benchmark Plots

Source: https://github.com/minishlab/semble/blob/main/benchmarks/README.md

Generate plots visualizing benchmark results, such as speed vs. NDCG for cold and warm starts.

```bash
uv run python -m benchmarks.plot
```

--------------------------------

### Run CodeRankEmbed Benchmark

Source: https://github.com/minishlab/semble/blob/main/benchmarks/README.md

Benchmark using CodeRankEmbed. Requires the 'benchmark' extra to be installed (`uv sync --extra benchmark`). Supports different modes, including 'semantic'.

```bash
uv run python -m benchmarks.baselines.coderankembed
```

```bash
uv run python -m benchmarks.baselines.coderankembed --mode semantic
```

--------------------------------

### Run Ripgrep Benchmark

Source: https://github.com/minishlab/semble/blob/main/benchmarks/README.md

Benchmark using 'ripgrep' (`rg`). Ensure 'rg' is installed on your system. Supports disabling fixed string matching.

```bash
uv run python -m benchmarks.baselines.ripgrep
```

```bash
uv run python -m benchmarks.baselines.ripgrep --no-fixed-strings
```

--------------------------------

### Initialize Semble Sub-agent

Source: https://github.com/minishlab/semble/blob/main/README.md

Run `semble init` to set up a Semble search sub-agent for various tools. The command creates configuration files in tool-specific directories. Prefix with `uvx --from "semble[mcp]"` if Semble is not on PATH.

```bash
semble init                      # Claude Code  → .claude/agents/semble-search.md
```

```bash
semble init --agent gemini       # Gemini CLI   → .gemini/agents/semble-search.md
```

```bash
semble init --agent cursor       # Cursor       → .cursor/agents/semble-search.md
```

```bash
semble init --agent opencode     # OpenCode     → .opencode/agents/semble-search.md
```

```bash
semble init --agent copilot      # Copilot CLI  → .github/agents/semble-search.md
```

```bash
semble init --agent kiro         # Kiro         → .kiro/agents/semble-search.md
```

--------------------------------

### Index a local directory with Semble Python library

Source: https://github.com/minishlab/semble/blob/main/README.md

Initialize a Semble index from a local project directory using `SembleIndex.from_path()`. This prepares the project for programmatic search.

```python
from semble import SembleIndex

# Index a local directory
index = SembleIndex.from_path("./my-project")
```

--------------------------------

### Run Semble Benchmark

Source: https://github.com/minishlab/semble/blob/main/benchmarks/README.md

Execute the main Semble benchmark. You can specify individual repositories or languages to test.

```bash
uv run python -m benchmarks.run_benchmark
```

```bash
uv run python -m benchmarks.run_benchmark --repo fastapi --repo axios
```

```bash
uv run python -m benchmarks.run_benchmark --language python
```

--------------------------------

### Invoke Semble via Bash (Alternative to MCP)

Source: https://github.com/minishlab/semble/blob/main/README.md

If Semble is not on your PATH, use `uvx --from "semble[mcp]" semble` to invoke it. This method is necessary for sub-agent support and can be used alongside MCP.

```bash
uvx --from "semble[mcp]" semble search "authentication flow" ./my-project
```

```bash
uvx --from "semble[mcp]" semble find-related src/auth.py 42 ./my-project
```

--------------------------------

### Run Context-Efficiency Benchmark

Source: https://github.com/minishlab/semble/blob/main/benchmarks/README.md

Benchmark context efficiency, measuring recall against token budget. Requires the 'benchmark' extra and 'rg` on PATH. Commands are provided for generating recall data and plotting results.

```bash
# Recall vs. token-budget across all queries; plots automatically.
uv run python -m benchmarks.token_efficiency recall
```

```bash
uv run python -m benchmarks.token_efficiency recall --repo fastapi
```

```bash
# Regenerate the plot from a saved recall payload.
uv run python -m benchmarks.token_efficiency plot
```

--------------------------------

### Run GrepAI Benchmark

Source: https://github.com/minishlab/semble/blob/main/benchmarks/README.md

Benchmark using 'grepai'. Requires 'grepai' on PATH and Ollama running with 'nomic-embed-text'. Supports specifying repositories and setting a timeout for large repos. The `--output` flag enables resume mode.

```bash
ollama pull nomic-embed-text
```

```bash
uv run python -m benchmarks.baselines.grepai
```

```bash
uv run python -m benchmarks.baselines.grepai --repo fastapi --repo axios
```

```bash
uv run python -m benchmarks.baselines.grepai --timeout 1800 --output results.json
```

--------------------------------

### Configure Semble Context Server for Zed

Source: https://github.com/minishlab/semble/blob/main/README.md

Add Semble context server configuration to Zed's settings.json file. This enables Zed to use Semble for context.

```json
{
  "context_servers": {
    "semble": {
      "command": "uvx",
      "args": ["--from", "semble[mcp]", "semble"]
    }
  }
}
```

--------------------------------

### Run ColGREP Benchmark

Source: https://github.com/minishlab/semble/blob/main/benchmarks/README.md

Benchmark using the 'colgrep' binary. Requires 'colgrep' on PATH. Allows specifying repositories. Note the `--code-only` flag behavior with different repository types.

```bash
uv run python -m benchmarks.baselines.colgrep
```

```bash
uv run python -m benchmarks.baselines.colgrep --repo fastapi --repo axios
```

--------------------------------

### Clone and Update Repositories

Source: https://github.com/minishlab/semble/blob/main/benchmarks/README.md

Use these commands to clone or update repositories listed in `repos.json`. The `--check` flag verifies the integrity of the cloned repositories.

```bash
uv run python -m benchmarks.sync_repos          # clone / update
```

```bash
uv run python -m benchmarks.sync_repos --check  # verify only
```

--------------------------------

### Run Speed Benchmark

Source: https://github.com/minishlab/semble/blob/main/benchmarks/README.md

Initiates the speed benchmark, which measures performance across different repositories and languages.

```bash
uv run python -m benchmarks.speed_benchmark
```

--------------------------------

### Search a local repository with Semble CLI

Source: https://github.com/minishlab/semble/blob/main/README.md

Use the `semble search` command to find code matching a query within a local project directory. The path defaults to the current directory if omitted.

```bash
semble search "authentication flow" ./my-project
```

--------------------------------

### Search Code by Description

Source: https://github.com/minishlab/semble/blob/main/src/semble/agents/semble-search.md

Use `semble search` to find code by describing its functionality or naming a symbol. Specify the search query and the target directory. The `--top-k` flag can limit the number of results.

```bash
semble search "authentication flow" ./my-project
```

```bash
semble search "save_pretrained" ./my-project
```

```bash
semble search "save model to disk" ./my-project --top-k 10
```

--------------------------------

### Index a remote Git repository with Semble Python library

Source: https://github.com/minishlab/semble/blob/main/README.md

Use `SembleIndex.from_git()` to create an index from a remote Git repository. Semble handles cloning and indexing for you.

```python
from semble import SembleIndex

# Index a remote git repository
index = SembleIndex.from_git("https://github.com/MinishLab/model2vec")
```

--------------------------------

### Configure Semble MCP Server for Codex

Source: https://github.com/minishlab/semble/blob/main/README.md

Add Semble MCP server configuration to Codex's config.toml file. This enables Codex to leverage Semble's capabilities.

```toml
[mcp_servers.semble]
command = "uvx"
args = ["--from", "semble[mcp]", "semble"]
```

--------------------------------

### Configure Semble MCP Server for OpenCode

Source: https://github.com/minishlab/semble/blob/main/README.md

Add Semble MCP server configuration to OpenCode's config.json file. This integrates Semble into OpenCode's workflow.

```json
{
  "mcp": {
    "semble": {
      "type": "local",
      "command": ["uvx", "--from", "semble[mcp]", "semble"]
    }
  }
}
```

--------------------------------

### Run Ablation Studies

Source: https://github.com/minishlab/semble/blob/main/benchmarks/README.md

Execute ablation studies to evaluate different components or configurations of Semble. Supports various modes like 'bm25' and 'semble-semantic'.

```bash
uv run python -m benchmarks.baselines.ablations
```

```bash
uv run python -m benchmarks.baselines.ablations --mode bm25
```

```bash
uv run python -m benchmarks.baselines.ablations --mode semble-semantic
```

--------------------------------

### Configure Semble MCP Server for VS Code

Source: https://github.com/minishlab/semble/blob/main/README.md

Add Semble MCP server configuration to VS Code's mcp.json file. This allows VS Code to utilize Semble.

```json
{
  "servers": {
    "semble": {
      "command": "uvx",
      "args": ["--from", "semble[mcp]", "semble"]
    }
  }
}
```

--------------------------------

### Search a remote repository with Semble CLI

Source: https://github.com/minishlab/semble/blob/main/README.md

Semble can search code in remote Git repositories. The repository will be cloned on demand. Git URLs are accepted.

```bash
semble search "save model to disk" https://github.com/MinishLab/model2vec
```

--------------------------------

### View Semble token savings summary

Source: https://github.com/minishlab/semble/blob/main/README.md

The `semble savings` command provides an overview of token savings across different periods. Use `--verbose` for a breakdown by call type.

```bash
semble savings
```

```bash
semble savings --verbose
```

--------------------------------

### Update Semble using pip or uv

Source: https://github.com/minishlab/semble/blob/main/README.md

Upgrade Semble to the latest version using pip or uv. For MCP users, clean the cache after upgrading and restart the MCP client.

```bash
pip install --upgrade semble
```

```bash
uv tool upgrade semble
```

```bash
uv cache clean semble
```

--------------------------------

### Configure Semble MCP Server for Cursor

Source: https://github.com/minishlab/semble/blob/main/README.md

Add Semble MCP server configuration to Cursor's mcp.json file. This allows Cursor to use Semble for code-related tasks.

```json
{
  "mcpServers": {
    "semble": {
      "command": "uvx",
      "args": ["--from", "semble[mcp]", "semble"]
    }
  }
}
```

--------------------------------

### Limit search results with Semble CLI

Source: https://github.com/minishlab/semble/blob/main/README.md

Control the number of results returned by `semble search` using the `--top-k` flag. This helps in managing the output for large codebases.

```bash
semble search "save model to disk" ./my-project --top-k 10
```

--------------------------------

### Search an index with Semble Python library

Source: https://github.com/minishlab/semble/blob/main/README.md

Perform searches against a Semble index using natural language or code queries. The `top_k` parameter limits the number of results.

```python
# Search the index with a natural-language or code query
results = index.search("save model to disk", top_k=3)
```

--------------------------------

### Search for a symbol or identifier with Semble CLI

Source: https://github.com/minishlab/semble/blob/main/README.md

Use `semble search` to locate specific symbols or identifiers within a project. This is useful for finding where a particular function or variable is used.

```bash
semble search "save_pretrained" ./my-project
```

--------------------------------

### Access search result details with Semble Python library

Source: https://github.com/minishlab/semble/blob/main/README.md

Each search result object in the Semble Python library provides access to the matched chunk's file path, line numbers, and content.

```python
# Each result exposes the matched chunk
result = results[0]
result.chunk.file_path   # "model2vec/model.py"
result.chunk.start_line  # 127
result.chunk.end_line    # 150
result.chunk.content     # "def save_pretrained(self, path: PathLike, ..."
```

--------------------------------

### Find Related Code

Source: https://github.com/minishlab/semble/blob/main/src/semble/agents/semble-search.md

Use `semble find-related` to discover code similar to a known location. Pass the file path and line number from a prior search result, along with the target directory.

```bash
semble find-related src/auth.py 42 ./my-project
```

--------------------------------

### Find related code with Semble Python library

Source: https://github.com/minishlab/semble/blob/main/README.md

Use `index.find_related()` to find code similar to a specific search result. This is useful for exploring code contextually.

```python
# Find code similar to a specific result
related = index.find_related(results[0], top_k=3)
```

=== COMPLETE CONTENT === This response contains all available snippets from this library. No additional content exists. Do not make further requests.