### Build and Setup MCP-Sophon Benchmarks

Source: https://github.com/lacausecrypto/mcp-sophon/blob/main/benchmarks/README.md

Builds the Rust project and sets environment variables for running benchmarks. Ensure you have Rust and Cargo installed.

```bash
cargo build --release -p mcp-integration
export SOPHON_BIN=./sophon/target/release/sophon
export SOPHON_REPO_ROOT=$(pwd)
```

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

### Recommended Sophon Runtime Setup

Source: https://github.com/lacausecrypto/mcp-sophon/blob/main/README.md

Set environment variables for persistent memory, on-disk retriever store, and BM25+Hash hybrid retrieval. Then, start the Sophon server.

```bash
# Persistent memory + on-disk retriever store + BM25+Hash hybrid
export SOPHON_MEMORY_PATH=~/.sophon/memory.jsonl
export SOPHON_RETRIEVER_PATH=~/.sophon/retriever
export SOPHON_HYBRID=1

sophon serve
```

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

### Install and verify mcp-sophon

Source: https://github.com/lacausecrypto/mcp-sophon/blob/main/npm/README.md

Install the package globally and verify the installation by running the help command.

```bash
npm install -g mcp-sophon
sophon --help
```

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

### Install MCP-Sophon via npm

Source: https://github.com/lacausecrypto/mcp-sophon/blob/main/README.md

Install the MCP-Sophon CLI globally using npm. Run 'sophon doctor' to verify the installation and check configuration.

```bash
npm install -g mcp-sophon
sophon doctor          # verify install + show config
```

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

### Install Sophon Python Package

Source: https://github.com/lacausecrypto/mcp-sophon/blob/main/sophon-py/README.md

Install the Sophon Python package using pip.

```bash
pip install sophon
```

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

### Sophon CLI Quick Commands

Source: https://github.com/lacausecrypto/mcp-sophon/blob/main/README.md

Execute common Sophon CLI commands for running tests, compressing prompts, installing agents, and viewing statistics.

```bash
sophon exec -- cargo test                       # run + compress combined output
sophon compress-prompt --prompt ./system.txt --query "rust errors" --max-tokens 500
sophon hook install --agent claude --global     # transparent Claude Code integration
sophon stats --period session                   # token savings rollup
```

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

### Build Sophon from Source

Source: https://github.com/lacausecrypto/mcp-sophon/blob/main/sophon-py/README.md

Build the Sophon Python package from source using maturin. This involves installing maturin, navigating to the directory, and running the development build command.

```bash
pip install maturin
cd sophon-py
maturin develop --release
pytest tests/
```

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

### Install Sophon Hook for Claude Agent

Source: https://github.com/lacausecrypto/mcp-sophon/blob/main/README.md

Installs Sophon transparently for the Claude agent globally. This is the primary method for integrating Sophon with code-focused LLM agents.

```bash
sophon hook install --agent claude --global
```

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

### Build with All Features

Source: https://github.com/lacausecrypto/mcp-sophon/blob/main/README.md

Build the mcp-integration binary with all available features, including tree-sitter and BGE embedder. This results in the largest binary size.

```bash
# All features (~42 MB):
cargo build --release -p mcp-integration --features "codebase-navigator/tree-sitter,bge"
```

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

### Run LLMLingua Head-to-Head Benchmark

Source: https://github.com/lacausecrypto/mcp-sophon/blob/main/benchmarks/README.md

Compares LLMLingua performance. Requires creating a data directory and potentially copying a system prompt file.

```bash
mkdir -p benchmarks/data
cp sophon/tests/fixtures/claude_system_prompt.txt benchmarks/data/system_prompt_large.txt  # or use your own
python3 benchmarks/llmlingua_compare.py
```

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

### Benchmark Comparison Scripts

Source: https://github.com/lacausecrypto/mcp-sophon/blob/main/CONTRIBUTING.md

Set up environment variables and run various Python scripts to compare Sophon's performance against other tools.

```bash
cargo build --release -p mcp-integration
export SOPHON_BIN=./sophon/target/release/sophon
export SOPHON_REPO_ROOT=$(pwd)

python3 benchmarks/llmlingua_compare.py   # § 7.8.d (Sophon vs LLMLingua-2)
python3 benchmarks/locomo_retrieval.py    # § 3.7 (Sophon 4-condition LOCOMO)
python3 benchmarks/locomo_mem0lite.py     # § 7.8.e (mem0-lite comparison)
python3 benchmarks/repos_scan.py          # § 7.2 / § 7.8.b (real-repo scan)
python3 benchmarks/repos_recall.py        # § 7.3 (recall@K)
```

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

### Build MCP-Sophon from Source

Source: https://github.com/lacausecrypto/mcp-sophon/blob/main/README.md

Clone the repository and build the mcp-integration binary from source. This requires Rust 1.75+.

```bash
git clone https://github.com/lacausecrypto/mcp-sophon
cd mcp-sophon/sophon
cargo build --release -p mcp-integration       # ~5.2 MB binary
```

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

### Build and Test Sophon

Source: https://github.com/lacausecrypto/mcp-sophon/blob/main/CONTRIBUTING.md

Build the release binary for the mcp-integration crate and run the entire test suite.

```bash
git clone https://github.com/lacausecrypto/mcp-sophon
cd mcp-sophon/sophon
cargo build --release -p mcp-integration
cargo test --workspace
```

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

### Run Repository QA Benchmark

Source: https://github.com/lacausecrypto/mcp-sophon/blob/main/BENCHMARK.md

Executes the repository question-answering benchmark script. Ensure the Sophon binary is built and accessible.

```bash
python3 benchmarks/repo_qa.py
```

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

### Run Session Token Economics Benchmark

Source: https://github.com/lacausecrypto/mcp-sophon/blob/main/BENCHMARK.md

Executes the session token economics benchmark script. Ensure the Sophon binary is built and accessible.

```bash
python3 benchmarks/session_token_economics.py
```

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

### Build Sophon Integration Binary

Source: https://github.com/lacausecrypto/mcp-sophon/blob/main/BENCHMARK.md

Builds the mcp-integration package in release mode for benchmarking. This command produces a 34 MB binary.

```bash
cargo build --release -p mcp-integration
```

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

### Run Prompt Compression Extended Benchmark

Source: https://github.com/lacausecrypto/mcp-sophon/blob/main/BENCHMARK.md

Executes the extended prompt compression benchmark script. Ensure the Sophon binary is built and accessible.

```bash
python3 benchmarks/prompt_compression_extended.py
```

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

### Run LOCOMO Cross-Model Benchmark

Source: https://github.com/lacausecrypto/mcp-sophon/blob/main/BENCHMARK.md

Executes the LOCOMO cross-provider A/B benchmark. Requires the LOCOMO dataset to be downloaded and its path set in SOPHON_BENCH_LOCOMO.

```bash
export SOPHON_BENCH_LOCOMO=/path/to/locomo/dir  # with all_items.jsonl
python3 benchmarks/locomo_cross_model.py      # cross-provider A/B
```

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

### Format Code

Source: https://github.com/lacausecrypto/mcp-sophon/blob/main/CONTRIBUTING.md

Format all code in the project according to the defined style.

```bash
cargo fmt --all
```

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

### Run All Tests

Source: https://github.com/lacausecrypto/mcp-sophon/blob/main/CONTRIBUTING.md

Execute all unit and integration tests across the workspace.

```bash
cargo test --workspace --no-fail-fast
```

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

### Build and Test Sophon Workspace

Source: https://github.com/lacausecrypto/mcp-sophon/blob/main/sophon/README.md

Commands to build and test the Sophon Rust workspace, including options for enabling tree-sitter for AST extraction.

```bash
# default build — regex extractors, no C deps, ~17 MB binary
car go build --release -p mcp-integration
car go test --workspace

# opt into 11-language AST extraction, ~25 MB binary
car go build --release -p mcp-integration --features codebase-navigator/tree-sitter
car go test --features codebase-navigator/tree-sitter

```

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

### Run Rust Test Suite

Source: https://github.com/lacausecrypto/mcp-sophon/blob/main/README.md

Execute the full Rust test suite for the mcp-sophon project, excluding the prompt-compressor feature.

```bash
cd sophon && cargo test --workspace --lib --tests --exclude prompt-compressor   # 405 tests
```

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

### Run Cross-Model Benchmarks (Default)

Source: https://github.com/lacausecrypto/mcp-sophon/blob/main/benchmarks/README.md

Executes the cross-model benchmark script with default settings. This configuration uses Stack A with a Sonnet answerer/judge and Stack B with a Codex answerer, while the internal Sophon LLM remains on Haiku.

```bash
python3 benchmarks/locomo_cross_model.py
```

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

### Build with Tree-sitter AST Extraction Feature

Source: https://github.com/lacausecrypto/mcp-sophon/blob/main/README.md

Build the mcp-integration binary with the 11-language tree-sitter AST extraction feature enabled. This increases the binary size.

```bash
# 11-language tree-sitter AST extraction (~25 MB):
cargo build --release -p mcp-integration --features codebase-navigator/tree-sitter
```

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

### Run LOCOMO Stratified Benchmark

Source: https://github.com/lacausecrypto/mcp-sophon/blob/main/BENCHMARK.md

Executes the LOCOMO benchmark with stratified sampling (N=30). Requires the LOCOMO dataset to be downloaded and its path set in SOPHON_BENCH_LOCOMO.

```bash
export SOPHON_BENCH_LOCOMO=/path/to/locomo/dir  # with all_items.jsonl
python3 benchmarks/locomo_v032_ab.py          # N=30 stratified
```

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

### Run Sophon Benchmarks

Source: https://github.com/lacausecrypto/mcp-sophon/blob/main/sophon/README.md

Commands to run benchmarks for specific Sophon crates using Criterion.

```bash
car go bench -p prompt-compressor
car go bench -p memory-manager
car go bench -p delta-streamer

```

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

### Build with BGE Semantic Embedder Feature

Source: https://github.com/lacausecrypto/mcp-sophon/blob/main/README.md

Build the mcp-integration binary with the BGE-small semantic embedder feature. Activate it at runtime using the SOPHON_EMBEDDER environment variable.

```bash
# BGE-small semantic embedder (~34 MB), activate with SOPHON_EMBEDDER=bge:
cargo build --release -p mcp-integration --features bge
```

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

### Run LOCOMO Retrieval Benchmark

Source: https://github.com/lacausecrypto/mcp-sophon/blob/main/benchmarks/README.md

Executes the LOCOMO retrieval benchmark. This requires a 'locomo' data directory and the 'all_items.jsonl' dataset.

```bash
mkdir -p benchmarks/data/locomo
# ...place all_items.jsonl here...
python3 benchmarks/locomo_retrieval.py
```

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

### Run Cross-Model Benchmarks (Focused)

Source: https://github.com/lacausecrypto/mcp-sophon/blob/main/benchmarks/README.md

Executes the cross-model benchmark script with a focus on adversarial items. This configuration limits the benchmark to 10 samples per type for the adversarial category.

```bash
CROSS_N=10 CROSS_TYPES=adversarial python3 benchmarks/locomo_cross_model.py
```

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

### Run mem0-lite LOCOMO Benchmark

Source: https://github.com/lacausecrypto/mcp-sophon/blob/main/benchmarks/README.md

Executes the mem0-lite LOCOMO benchmark. This is a faster variant of the LOCOMO retrieval benchmark.

```bash
python3 benchmarks/locomo_mem0lite.py
```

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

### Run Cross-Model Benchmarks (All-Claude)

Source: https://github.com/lacausecrypto/mcp-sophon/blob/main/benchmarks/README.md

Executes the cross-model benchmark script using only Claude models. Both Stack A and Stack B utilize Claude models, with Stack A using Sonnet and Stack B using Opus as answerers.

```bash
CROSS_ANSWERER_A="python3 benchmarks/llm_cli.py --provider claude --model sonnet" \
CROSS_ANSWERER_B="python3 benchmarks/llm_cli.py --provider claude --model opus" \
python3 benchmarks/locomo_cross_model.py
```

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

### Project Directory Structure

Source: https://github.com/lacausecrypto/mcp-sophon/blob/main/README.md

This snippet displays the directory layout of the MCP-Sophon project. It shows the main files and subdirectories, including the location of crates for different functionalities.

```bash
.
├── README.md           ← you are here
├── BENCHMARK.md        ← full per-section benchmark detail
├── CHANGELOG.md        ← version history + deprecated numbers
├── benchmarks/         ← reproducible scripts for every number above
├── npm/                ← npm wrapper package
└── sophon/crates/      ← 11-crate Rust workspace
    ├── prompt-compressor/    compress_prompt
    ├── memory-manager/       compress_history, update_memory, rolling summary
    ├── delta-streamer/       read/write_file_delta
    ├── fragment-cache/       encode/decode_fragments
    ├── semantic-retriever/   chunker + HashEmbedder + BM25 + entity graph
    ├── output-compressor/    21 command-aware filters + JsonStructural
    ├── codebase-navigator/   tree-sitter / regex + PageRank
    ├── cli-hooks/            transparent agent installer
    └── mcp-integration/      stdio server, async dispatch, cancellation

```

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

### Run Python Test Suite

Source: https://github.com/lacausecrypto/mcp-sophon/blob/main/README.md

Execute the Python test suite for the mcp-sophon project using pytest.

```bash
cd sophon-py && .venv/bin/pytest tests/                                         # 4 Python tests
```

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

### Set Sophon Binary Path

Source: https://github.com/lacausecrypto/mcp-sophon/blob/main/BENCHMARK.md

Sets the SOPHON_BIN environment variable to point to the compiled Sophon executable. This is a prerequisite for running most benchmark scripts.

```bash
export SOPHON_BIN=./sophon/target/release/sophon
```

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

### Run Rust AST Tests

Source: https://github.com/lacausecrypto/mcp-sophon/blob/main/README.md

Execute Rust tests that include Abstract Syntax Tree (AST) functionality, specifically for the codebase-navigator feature.

```bash
cd sophon && cargo test --features codebase-navigator/tree-sitter               # +AST tests
```

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

### Configure mcp-sophon as an MCP server

Source: https://github.com/lacausecrypto/mcp-sophon/blob/main/npm/README.md

Use this JSON configuration to set up Sophon as an MCP server within your project.

```json
{
  "mcpServers": {
    "sophon": {
      "command": "sophon",
      "args": ["serve"]
    }
  }
}
```

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

### Sophon Workspace Layout

Source: https://github.com/lacausecrypto/mcp-sophon/blob/main/sophon/README.md

Overview of the directory structure for the Sophon Rust workspace, detailing the purpose of each crate.

```text
sophon/
├── Cargo.toml
├── sophon.toml          runtime defaults
├── crates/
│   ├── sophon-core/          shared types, cl100k_base counting, hashing
│   ├── prompt-compressor/    compress_prompt
│   ├── memory-manager/       compress_history, update_memory, persistence
│   ├── delta-streamer/       read_file_delta, write_file_delta
│   ├── fragment-cache/       encode_fragments, decode_fragments
│   ├── semantic-retriever/   chunker + HashEmbedder + linear k-NN
│   ├── output-compressor/    command-aware stdout/stderr compression
│   ├── cli-hooks/            transparent command rewriter + installer
│   ├── codebase-navigator/   tree-sitter/regex extractors + PageRank + digest
│   └── mcp-integration/      MCP stdio server, CLI, tool schemas
└── tests/
    └── fixtures/

```

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

### Sophon Prompt Compression

Source: https://github.com/lacausecrypto/mcp-sophon/blob/main/sophon-py/README.md

Use Sophon's prompt compression feature to select relevant sections of a prompt based on a query. This helps in reducing token usage for large prompts.

```python
from sophon import Sophon

s = Sophon()

# Prompt compression (query-driven section selection)
result = s.compress_prompt(
    "<rust>use Result and ?</rust><web>fetch()</web>",
    "rust errors",
    max_tokens=500,
)
print(result.compressed_prompt)
print(f"{result.token_count} tokens, ratio={result.compression_ratio:.2f}")
```

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

### compress_prompt

Source: https://github.com/lacausecrypto/mcp-sophon/blob/main/sophon/README.md

Compresses a prompt by trimming less relevant sections while preserving topic-matched content, especially under a tight token budget.

```APIDOC
## compress_prompt

### Description
Compresses a prompt by trimming less relevant sections while preserving topic-matched content, especially under a tight token budget.

### Method
POST

### Endpoint
/tools/compress_prompt

### Parameters
#### Request Body
- **prompt** (string) - Required - The input prompt.
- **query** (string) - Required - The query used for topic matching.
- **max_tokens** (integer) - Optional - The maximum number of tokens allowed for the compressed prompt.
```

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

### Run Latency Reliability Benchmark

Source: https://github.com/lacausecrypto/mcp-sophon/blob/main/BENCHMARK.md

Executes the latency and reliability benchmark script. Sets environment variables LAT_N and LAT_N_HEAVY to control sample sizes.

```bash
LAT_N=30 LAT_N_HEAVY=10 python3 benchmarks/latency_reliability.py
```

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

### Agent Session Token Economics Breakdown

Source: https://github.com/lacausecrypto/mcp-sophon/blob/main/BENCHMARK.md

This table breaks down token usage and savings per operation within a simulated 25-turn coding session. It shows the raw tokens, sent tokens, and percentage saved for each operation, illustrating Sophon's efficiency.

```text
Operation | Count | Raw tokens | Sent tokens | Saved |
|---|---|---|---|---|
| `compress_output` | 5 | 11 480 | 2 320 | **79.8 %** |
| `write_file_delta` | 3 | 4 644 | 258 | **94.4 %** |
| `file_read_baseline` | 3 | 4 599 | 4 599 | 0.0 % |
| `read_file_delta` (re-reads) | 3 | 4 599 | 90 | **98.0 %** |
| `compress_history` | 9 | 3 615 | 2 483 | 31.3 % |
| `compress_prompt` | 2 | 1 900 | 88 | **95.4 %** |
| **TOTAL** | **25** | **30 837** | **9 838** | **68.1 %** |
```

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

### LOCOMO Benchmark Confidence Intervals

Source: https://github.com/lacausecrypto/mcp-sophon/blob/main/BENCHMARK.md

Displays the 95% Wilson Confidence Intervals for LOCOMO benchmark results across different system versions. Overlapping CIs indicate results are not statistically robust.

```text
V030       33.3 %   [19.2 — 51.2]
V032_FULL  40.0 %   [24.6 — 57.7]
FULL       73.3 %   [55.6 — 85.8]
```

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

### Programmatic JSON-RPC Call to Sophon

Source: https://github.com/lacausecrypto/mcp-sophon/blob/main/README.md

Send a one-shot JSON-RPC request to the Sophon server to perform an action, such as compressing a prompt with specific arguments.

```bash
echo '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"compress_prompt","arguments":{"prompt":"<rust>?: operator</rust><web>fetch()</web>","query":"rust errors","max_tokens":500}}}' \
  | sophon serve
```

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

### Sophon Output Compression

Source: https://github.com/lacausecrypto/mcp-sophon/blob/main/sophon-py/README.md

Utilize Sophon's output compression to filter long command outputs. This is useful for commands that produce verbose results, like test suites.

```python
# Output compression (command-aware filtering)
out = s.compress_output("cargo test", long_test_output)
print(out.compressed)
print(f"ratio={out.ratio:.2f}, filter={out.filter_name}")
```

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

### Compress History Configuration Flags

Source: https://github.com/lacausecrypto/mcp-sophon/blob/main/BENCHMARK.md

These environment variables are used to activate the v0.4.0 flag stack for compress_history, significantly impacting its performance and reliability.

```shell
SOPHON_HYDE=1  SOPHON_FACT_CARDS=1  SOPHON_ADAPTIVE=1
SOPHON_LLM_RERANK=1  SOPHON_TAIL_SUMMARY=1
SOPHON_ENTITY_GRAPH=1  SOPHON_CHUNK_TARGET=500
```

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

### get_token_stats

Source: https://github.com/lacausecrypto/mcp-sophon/blob/main/sophon/README.md

Retrieves cumulative per-module token statistics, including compression ratios.

```APIDOC
## get_token_stats

### Description
Retrieves cumulative per-module token statistics, including compression ratios.

### Method
GET

### Endpoint
/tools/get_token_stats

### Response
#### Success Response (200)
- **stats** (object) - An object containing per-module statistics.
  - **compression_ratio** (number) - The compression ratio (compressed / original), clamped to [0, 1].
```

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

### update_memory

Source: https://github.com/lacausecrypto/mcp-sophon/blob/main/sophon/README.md

Manages stateful session memory by appending new information and creating snapshots without resending the full history. Can persist to disk.

```APIDOC
## update_memory

### Description
Manages stateful session memory by appending new information and creating snapshots without resending the full history. Can persist to disk.

### Method
POST

### Endpoint
/tools/update_memory

### Parameters
#### Request Body
- **state** (object) - Required - The current memory state.
- **append** (object) - Required - New information to append to the memory.

### Notes
Set `SOPHON_MEMORY_PATH=~/.sophon/memory/session.jsonl` to persist appends to disk across server restarts.
```

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

### Sophon Token Counting

Source: https://github.com/lacausecrypto/mcp-sophon/blob/main/sophon-py/README.md

Count the number of tokens in a given string using the cl100k_base encoding with Sophon's static method.

```python
# Token counting (cl100k_base)
tokens = Sophon.count_tokens("hello world")
```

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

### compress_history

Source: https://github.com/lacausecrypto/mcp-sophon/blob/main/sophon/README.md

Compresses conversation history. Short histories that already fit within the budget are passed through unchanged.

```APIDOC
## compress_history

### Description
Compresses conversation history. Short histories that already fit within the budget are passed through unchanged.

### Method
POST

### Endpoint
/tools/compress_history

### Parameters
#### Request Body
- **messages** (array) - Required - An array of message objects representing the conversation history.
- **max_tokens** (integer) - Optional - The maximum number of tokens allowed for the compressed history.
- **recent_window** (integer) - Optional - The number of recent messages to prioritize.
- **include_index** (boolean) - Optional - Whether to include an index in the compressed history (default: false).
```

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

### write_file_delta

Source: https://github.com/lacausecrypto/mcp-sophon/blob/main/sophon/README.md

Performs an atomic staged write of file content or delta.

```APIDOC
## write_file_delta

### Description
Performs an atomic staged write of file content or delta.

### Method
POST

### Endpoint
/tools/write_file_delta

### Parameters
#### Request Body
- **path** (string) - Required - The path to the file.
- **content** (string) - Required - The file content or delta to write.
```

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

### count_tokens

Source: https://github.com/lacausecrypto/mcp-sophon/blob/main/sophon/README.md

Calculates the ground-truth token count for a given string using the cl100k_base encoding.

```APIDOC
## count_tokens

### Description
Calculates the ground-truth token count for a given string using the cl100k_base encoding.

### Method
POST

### Endpoint
/tools/count_tokens

### Parameters
#### Request Body
- **text** (string) - Required - The input text to count tokens for.
```

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

### read_file_delta

Source: https://github.com/lacausecrypto/mcp-sophon/blob/main/sophon/README.md

Reads a file delta, returning the status of the file content relative to a known version or hash.

```APIDOC
## read_file_delta

### Description
Reads a file delta, returning the status of the file content relative to a known version or hash.

### Method
POST

### Endpoint
/tools/read_file_delta

### Parameters
#### Request Body
- **path** (string) - Required - The path to the file.
- **known_version** (string) - Optional - The known version of the file.
- **known_hash** (string) - Optional - The known hash of the file.

### Response
#### Success Response (200)
- **status** (string) - 'Unchanged', 'Delta', or 'Full' - Indicates the file status.
- **content** (string) - The file content or delta (if applicable).
```

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

### decode_fragments

Source: https://github.com/lacausecrypto/mcp-sophon/blob/main/sophon/README.md

Reconstructs a document from encoded fragments, replacing fragment identifiers with their original content.

```APIDOC
## decode_fragments

### Description
Reconstructs a document from encoded fragments, replacing fragment identifiers with their original content.

### Method
POST

### Endpoint
/tools/decode_fragments

### Parameters
#### Request Body
- **encoded_document** (string) - Required - The document with encoded fragments.
- **fragments** (object) - Required - A mapping of fragment identifiers to their original content.
```

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

### encode_fragments

Source: https://github.com/lacausecrypto/mcp-sophon/blob/main/sophon/README.md

Deduplicates repeated multi-paragraph blocks within a document by encoding them into fragments.

```APIDOC
## encode_fragments

### Description
Deduplicates repeated multi-paragraph blocks within a document by encoding them into fragments.

### Method
POST

### Endpoint
/tools/encode_fragments

### Parameters
#### Request Body
- **document** (string) - Required - The input document.
- **max_window** (integer) - Optional - The maximum window size for detector (defaults to 64 paragraphs, or paragraphs/2).

### Notes
The detector window is adaptive and can be overridden with `SOPHON_FRAGMENT_MAX_WINDOW`.
```

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