doclex-cli 0.1.1

Local document search — BM25 lexical search over PDFs, DOCX, XLSX, CSV, Markdown and TXT files. No embeddings, no vector DB, no cloud.

Doclex

English · Español

A CLI that turns a local folder of documents into a searchable, citable corpus. Point it at a folder; extraction, segmentation, indexing, and change tracking happen automatically.

Features

Doclex is the orchestration layer that ties them together so you never have to think about the pipeline yourself:

• Plug and play — point it at a folder. Extraction, structural segmentation, indexing, and change tracking all happen automatically; there's no pipeline to wire up.
• SQLite is the source of truth — every file's hash, mtime, and indexing status live in one place, so Doclex always knows exactly what changed and reprocesses only that.
• Local-first — everything runs on your machine, nothing leaves it. No embeddings, no vector database, no cloud.
• Deterministic — BM25 lexical scoring, no black-box similarity.
• Traceable — every result cites its file, page/sheet/section, and score.
• Incremental — only reprocesses files that actually changed.

Install

Requires Python ≥3.12 and uv.

git clone https://github.com/EnPaiva93/doclex-cli.git
cd doclex-cli
uv sync

Quickstart

# register a folder as a project
uv run doclex add mydocs ./path/to/documents

# build the index
uv run doclex index --project mydocs

# search it
uv run doclex search "quarterly budget" --project mydocs

[1] budget_2024.pdf — page 3  (score: 4.812)
    "...the quarterly budget was approved with a 12% increase in..."

[2] notes.md — Planning  (score: 2.107)
    "...need to revisit the budget assumptions before..."

Walkthrough

A typical day-to-day flow, from an empty project to keeping it in sync:

# 1. register the folder you want to search
doclex add contracts ~/Documents/legal/contracts

# 2. build the index — extracts, segments, and indexes every supported file
doclex index --project contracts

Processing: 42 new, 0 modified, 0 deleted
✓ Done.

# 3. search it
doclex search "termination clause" --project contracts

[1] MSA_2023_AcmeCorp.pdf — Section 9.2  (score: 6.104)
    "...either party may terminate this Agreement upon 30 days written notice..."

[2] vendor_agreement_v3.docx — Termination  (score: 5.887)
    "...termination for convenience requires 60 days advance notice to..."

# 4. check what's indexed
doclex status --project contracts

Folder       ~/Documents/legal/contracts
Indexed      42 files
Segments     318
BM25 index   318 docs
Disk state   up to date

# 5. a new contract shows up on disk — reindexing picks it up automatically
doclex index --project contracts

Processing: 1 new, 0 modified, 0 deleted
✓ Done.

# 6. see every registered project
doclex list

Name        Folder
contracts   /home/you/Documents/legal/contracts

# 7. done with it — deregister and purge its cache/index
doclex remove contracts

Each project is independent — register as many non-overlapping folders as you need, each with its own cache and index:

doclex add hr-policies ~/Documents/hr
doclex index --project hr-policies
doclex search "vacation days" --project hr-policies --top 5

Projects

Every command operates on a registered project — a name bound to a folder — instead of a raw --folder flag. This keeps folder resolution unambiguous and lets Doclex reject overlapping folders (a subfolder registered as its own project would silently duplicate indexing and caching):

doclex add mydocs ./path/to/documents
doclex add mydocs ./other       # error: name already exists
doclex add sub ./path/to/documents/subdir   # error: overlaps with 'mydocs'
doclex remove mydocs            # deregisters and purges its cache/index

Commands

Command	Description
doclex add <name> <folder>	Register a folder under a project name
doclex list	List registered projects
doclex remove <name>	Remove a project and purge its cache/index
doclex index --project <name>	Scan and index (incremental by default)
doclex search <query> --project <name>	Search and print results with citations
doclex status --project <name>	Show corpus/index status

Run doclex <command> --help for all flags and examples.

Flags in practice

# force a full re-extraction
doclex index --project mydocs --force

# reindex specific files after editing them
doclex index --project mydocs --file report.pdf --file notes.docx

# more results, full segment text instead of a snippet
doclex search "risk assessment" --project mydocs --top 20 --full

# list every indexed file and its status
doclex status --project mydocs --all --filter error

Architecture

folder → scan/watch → extract (Kreuzberg) → segment → BM25X index
                                                  ↕
                                          SQLite (files, segments)

Layer	Responsibility
Ingestion	Scans a project's folder, hashes files to detect new/modified/deleted ones, skips its own cache directory
Extraction	Delegates to Kreuzberg to turn each file into plain text, normalized to an internal format
Segmentation	Splits text along the file's natural structure — never blind fixed-size chunking (PDF → page, XLSX → sheet, DOCX → section, TXT/MD → block)
Index	Indexes segments (not whole files) with BM25X; supports incremental add/update/delete
Metadata	SQLite tracks each file's hash, mtime, size, MIME type, and indexing status, plus which segments belong to it

A local cache stores extracted text and generated segments per file, so unchanged files are never reprocessed.

Supported file types

PDF · TXT · Markdown · DOCX · XLSX · CSV

Design principles

Doclex is built to be local-first, incremental, explainable, deterministic, and cheap to run. It deliberately does not:

• use embeddings or a vector database
• upload files anywhere
• depend on external services to search
• run OCR by default
• generate answers — it retrieves evidence, it doesn't interpret it

Stack

Python · Typer · Kreuzberg · BM25X · Watchdog · SQLite

Development

uv sync --extra dev
uv run pytest