New Modules

• jsonx (v1.0.0): renamed from jsonc, now an extended JSON parser supporting JSONC (comments + trailing commas) and JSONL/NDJSON. New API: loads_lines, load_lines, dumps_lines, dump_lines. Batch fast-path matches ndjson speed on clean JSONL. (#95, #99)

New Features

• readability (v0.2.0): Expose best-candidate score in ReadabilityResult (#97). Extended score propagation from 2 to 4 ancestor levels for deeply nested SPA pages (#98).

Enhancements

• config: updated internal dependency from jsonc to jsonx with backward-compatible fallback to legacy jsonc module.
• CLI: added replaced_by mechanism — zerodep add jsonc now installs jsonx with a migration notice; zerodep outdated detects local jsonc.py and suggests upgrading.
• CLI: new modules now start at version 0.0.0 (sentinel); version-check --strict blocks release if any module is unbumped.

Bug Fixes

• httpserver: use chunked transfer encoding for SSE streaming responses (#101).
• CLI: fixed cmd_bump crash when no modules were bumped (empty rows in table formatting).

Performance

• jsonx: loads_lines uses batch parsing (join lines into JSON array, single json.loads call) for clean JSONL, achieving ndjson-level performance (10 lines: 5.1 μs vs ndjson 6.6 μs). Falls back to per-line JSONC processing when comments are present.

Infrastructure

• Release workflow: replaced auto-bump with version-check --strict gate — releases are blocked if any module has content changes without a version bump.
• Sync Readability Patterns workflow: fixed to use master branch (readability module is no longer on a feature branch).

Migration

Downstream repos using jsonc module should update:

• _vendor/jsonc.py → _vendor/jsonx.py
• from jsonc import loads → from jsonx import loads
• Or simply run zerodep add jsonc — it now installs jsonx automatically with a migration notice.

Full Changelog: v2026.6.1...v2026.6.13

What's Changed

• bench: add real-world fixture data for parser modules (#83) by @Oaklight in #88
• feat(bench): expose min/max/stddev/P95 in benchmark reports by @Oaklight in #89
• feat(bench): add tracemalloc memory benchmarks for parser modules by @Oaklight in #90
• feat(bench): geometric scale curves for yaml/soup/protobuf/multipart by @Oaklight in #91
• fix(bench): pair scale-curve tests by matching parametrize suffix by @Oaklight in #92
• fix(httpclient): URL-encode dict data without files by @Oaklight in #94

Full Changelog: v2026.5.17...v2026.6.1

Contributors

• @Oaklight

New Modules

• multipart: zero-dependency multipart/form-data parser and encoder (RFC 7578 / RFC 2046). parse_multipart(), encode_multipart(), extract_boundary() with Part dataclass. Boundary-split algorithm using bytes.find() for C-level performance — 3-7x faster than python-multipart. Robust boundary detection, Content-Transfer-Encoding (base64/quoted-printable), RFC 5987 filename* support. Security limits: max_part_size (10 MB), max_parts (1000). 57 correctness tests + 11 benchmarks.

Bug Fixes

• httpclient: Remove redundant global asyncio.Lock() / threading.Lock() from AsyncClient / Client that serialized concurrent requests and could cause deadlocks. Connection pools already provide internal fine-grained locking. (#76)

Infrastructure

• Added docs/changelog update workflow guidance to AGENTS.md.

Full Changelog: v2026.5.5.3...v2026.5.17

jsonschema: Generic JSON Pointer $ref Resolution (RFC 6901)

resolve_refs() now handles any local JSON Pointer fragment, not just #/$defs/ and #/definitions/. This enables resolving OpenAPI-style references like #/components/schemas/XXX, making it a drop-in replacement for jsonref.replace_refs().

Also adds circular reference protection to prevent infinite recursion on self-referencing schemas.

Module Version Bumps

Full Changelog: v2026.5.5.2...v2026.5.5.3

Module Version Bumps

Corrects the version for the FieldValidator and model_validator addition — new feature warrants a minor bump, not patch.

Full Changelog: v2026.5.5.1...v2026.5.5.2

Highlights

httpclient: SOCKS5 Proxy Support (#72)

Native SOCKS5 proxy support (RFC 1928) with username/password authentication (RFC 1929), implemented directly in the module with zero external dependencies.

from httpclient import get, Client
# SOCKS5 proxy
r = get("https://api.example.com/data", proxy="socks5://proxy:1080")
# With authentication
r = get("https://api.example.com/data", proxy="socks5://user:pass@proxy:1080")
# Session-level
with Client(proxy="socks5://proxy:1080") as c:
 r = c.get("https://api.example.com/data")

• Both sync (Client) and async (AsyncClient) supported
• HTTPS over SOCKS5 tunnel works out of the box
• Streaming responses work through SOCKS5
• ~6-15x faster than httpx[socks] in benchmarks

CLI Bug Fixes

• _list_release_tags() now filters legacy project-level SemVer tags (v0.x.y) that overlap with per-module SemVer and caused false content-hash comparisons in _find_changed_modules().
• Version-change detection uses != instead of > so that intentional version reverts are not misclassified as "modified".

Module Version Bumps

What's Changed

• feat(httpclient): add SOCKS5 proxy support by @Oaklight in #73

Full Changelog: v2026.5.2.1...v2026.5.5.1

Contributors

• @Oaklight

Hotfix for v2026.5.2.

Fixes

• CLI: version-check and bump no longer incorrectly auto-bump brand-new modules that have no prior release tag. New modules now show "new (initial release)" instead of "new (needs version bump)".
• Module versions: Reverted incorrect version bump on 5 new modules — httpserver (0.1.0), websocket (0.1.0), cdp (0.1.0), useragent (0.1.0), synctex (0.2.0).

New Modules

• websocket (0.1.0) — Zero-dependency RFC 6455 WebSocket client with sync + async support. WebSocketClient and AsyncWebSocketClient for ws:// and wss:// connections. Full protocol: upgrade handshake, text frame encoding/decoding with client-side masking, ping/pong auto-response, close frame exchange. TLS with optional cert verification. ~1000 lines, stdlib only. 27 correctness tests + 8 benchmarks (vs websockets).
• cdp (0.1.0) — Zero-dependency Chrome DevTools Protocol client for headless browser automation. High-level API: get_rendered_text() / get_rendered_html() for one-call SPA content extraction. Low-level API: target management, navigation, JS evaluation, user-agent spoofing. Auto-discovers browser debugger WebSocket URL. Depends on sibling websocket module. ~900 lines, stdlib only. 18 correctness tests + 10 benchmarks.
• httpserver (0.1.0) — Zero-dependency async HTTP server with Flask/Microdot-compatible decorator routing. @app.route(), @app.get(), @app.post() etc. Path parameters with type conversion. Return value coercion (dict → JSON, str → text, tuple → status/headers). StreamingResponse for SSE, FileResponse for static files. Middleware hooks, graceful shutdown. ~1000 lines, stdlib only. 58 correctness tests + 26 benchmarks (vs Flask, microdot, bottle).
• useragent (0.1.0) — Lightweight Chrome/Edge User-Agent string generator with Client Hints headers. Realistic UA strings for Windows, macOS, Linux (desktop) and Android (mobile). Deterministic output via random.seed(). ~470 lines. 47 correctness tests + 8 benchmarks (2-3x faster than ua-generator).
• synctex (0.2.0) — Zero-dependency SyncTeX bidirectional search parser (PDF ↔ source mapping). parse_synctex(), inverse_search(), forward_search(). Supports configurable path prefix stripping for Docker/remote builds. 29 correctness tests.

Enhancements

• CLI: zerodep new --from FILE — create modules from existing Python source files with automatic frontmatter injection
• sparse_search benchmarks: Added bm25s comparison and Zipf-distributed scale tests
• httpserver benchmarks: Concurrent serving (10 simultaneous requests), sync vs async handler overhead, large payload (~30KB JSON); Flask added as 4th benchmark target

Infrastructure

• Automated PyPI publish workflow on GitHub Release
• AGENTS.md for unified AI coding assistant instructions (Claude Code, Codex, Cursor, Copilot, Gemini CLI)
• Live benchmark dashboard badge added to README and docs

Full changelog: https://zerodep.readthedocs.io/en/latest/changelog/

Highlights

New Modules

• llmstxt — llms.txt parser and markdown URL candidate finder, with discover() for site-level probing
• png — PNG/BMP image codec with matrix compression API

New Features

• markdown 0.4.0 — GFM strikethrough (~~text~~), task lists (- [ ] / - [x]), extended autolinks
• qr 0.3.2 — qr_to_png() and qr_to_svg() image output (depends on new png module)
• validate 0.4.2 — FieldValidator for transform+validate, model_validator for cross-field validation
• sparse_search 0.4.0 — RRF (Reciprocal Rank Fusion) and MMR (Maximal Marginal Relevance) post-retrieval re-ranking
• soup 0.6.0 — CSS pseudo-selector support
• protobuf 0.4.3 — byte_size() method for size calculation without serialization

Performance

• protobuf — varint fast-paths, write-to-buffer encoders, _is_default_value specialization, encoder dispatch binding, map entry type caching
• png — optimized BMP codec, PNG filters, and mode conversion

Infrastructure

• Refine module categories from 7 to 12 fine-grained groups: network, protocol, serialization, validation, text, config, terminal, crypto, image, process, storage, devtools
• Rename scripts/ → _scripts/ (internal convention)
• Auto-generate modules/index.md during docs build from manifest.json + config
• Fix version-check false positives from old project-level SemVer tag collisions and wrong primary file selection

What's Changed

• perf(protobuf): optimize encode and decode hot paths by @Oaklight in #53
• perf(protobuf): optimize encode/decode hot paths and add byte_size() by @Oaklight in #54
• feat(soup): add CSS pseudo-selector support by @Oaklight in #55
• feat(markdown): add GFM strikethrough, task lists, and extended autolinks by @Oaklight in #56
• feat(png): add PNG/BMP image codec with matrix compression API by @Oaklight in #57
• perf(png): optimize BMP codec, PNG filters, and mode conversion by @Oaklight in #58
• feat(qr): add PNG and SVG image output by @Oaklight in #59
• feat(validate): add FieldValidator and model_validator by @Oaklight in #60
• feat(sparse_search): add RRF and MMR post-retrieval re-ranking by @Oaklight in #61
• feat(llmstxt): add llms.txt parser and markdown URL candidate finder by @Oaklight in #63
• feat(llmstxt): add discover() for site-level llms.txt probing by @Oaklight in #64

Full Changelog: v2026.4.25...v2026.4.27

Contributors

• @Oaklight

New Modules

• JSON Schema module: zero-dependency JSON Schema flattening & sanitization. flatten_schema() resolves $ref, merges allOf (deep-merge with type intersection, numeric constraint tightening, required union), simplifies anyOf/oneOf (nullable detection, single-variant unwrap), and strips unsupported keywords. Each phase is independently callable: resolve_refs, merge_allof, simplify_unions, sanitize. Benchmarked against allof-merge (JS) across 5 complexity tiers — 1.7-5.3x faster. 57 unit tests + 13 cross-implementation correctness tests.
• Readability module: zero-dependency article content extractor ported from Mozilla Readability.js. extract() performs full article extraction with metadata (title, author, excerpt, published_time, site_name, lang, dir). is_probably_readable() provides a quick readability heuristic check. Supports JSON-LD and OpenGraph metadata extraction. 2-level retry (ruthless on/off) for robust extraction. 18 Mozilla Readability.js test fixtures for cross-validation. Three-way benchmark: zerodep vs readability-lxml vs Mozilla JS. Depends on soup module (no pip dependencies).

Enhancements

• Soup module: extended with tree mutation and serialization APIs — append(), insert(), extract(), replace_with(), unwrap() for tree manipulation; to_html() / __str__() for HTML serialization; __setitem__ / __delitem__ for attribute setting/deletion; Soup.new_tag() factory method for creating detached Tag nodes.

Performance

• A2A module: optimized serialization and deserialization paths — serialization is now 2.2-2.6x faster than a2a-protocol (previously 1.4-2.1x slower). Deserialization narrowed from 1.8-4.1x slower to 1.1-1.7x slower. JSON round-trip is now 1.4-1.5x faster end-to-end.
• Cache module: optimized TypedKey construction, decorator wrapper path, and LFU eviction. LRU decorator overhead is now 1.2x faster than cachetools. LFU eviction now ~equal (previously 1.4x slower).
• PersistDict module: optimized SQLite write performance — PRAGMA synchronous=NORMAL, commit_every parameter for batched writes.
• QR module: optimized encoding path — short inputs now only 1.1x slower (previously 2.1x). Medium/long inputs now 1.1-1.2x faster (previously 1.7-1.9x slower).
• Readability module: optimized scoring and tree-walking — small pages 2.1x faster, medium pages now 1.2x faster (previously 2x slower). Large pages improved from 2x slower to 1.4x slower vs readability-lxml.
• Semver module: optimized parse, comparison, and property access. Parse ~1.3x faster, sort ~1.4x faster, compare ~1.4x faster, property access ~4.5x faster.

Bug Fixes

• Docs: fixed ReadTheDocs build failure caused by search → sparse_search module rename.
• zerodep CLI: fixed bump command double-bumping modules whose version was already incremented before tagging.

Full Changelog: v2026.4.15.1...v2026.4.25