udsdoc

AUTOSAR DCM/CanTp/DEM ARXML → UDS (ISO 14229) specification documents (LaTeX → PDF) and ASAM ODX 2.2 / PDX (ISO 22901-1).

udsdoc turns AUTOSAR diagnostic ARXML into a complete UDS specification and into ODX/PDX for diagnostic tooling. It offers two front-ends over the same pipeline: a GUI (local browser app) and a CLI (headless, CI/HPC-friendly).

Install

pip install udsdoc          # CLI core (LaTeX / Markdown / RST / ODX / PDX)
pip install udsdoc[web]     # GUI  (udsdoc --serve)
pip install udsdoc[all]     # everything (GUI + LLM verify + ECU testing)

PDF output needs a LaTeX distribution with xelatex (TeX Live / MiKTeX). No pip on the target machine? Grab the self-contained udsdoc.pyz from Releases — it runs with just a Python interpreter (python3 udsdoc.pyz input.arxml -o spec.tex).

GUI version

udsdoc --serve            # opens http://127.0.0.1:8765

The no-install static browser edition also generates LaTeX/PDX and performs local compliance analysis entirely in the tab. Its Pyodide runtime, Python wheels, WebLLM runtime, and odx-rs WASM are self-hosted and available offline after the first successful load.

Inputs — on the Generate tab, pick a BSW repository folder (or a built-in sample). It auto-discovers the ARXML (Dcm / CanTp / Dem) and the C sources; attach an optional Compliance Matrix CSV (+ a sources YAML) for a requirements-compliance chapter.
Outputs — Generate & Download a LaTeX ZIP that includes a compiled spec.pdf (Tex → PDF), or Export ODX / PDX for CANalyzer / CANoe.
LLM — the LLM tab connects the provider used by Verify Conditionally-Accepted comments: Claude (browser sign-in), ChatGPT / Gemini (paste an API key — the button opens the key page for you), or local Ollama (pick or install an open-source model). Credentials are stored locally under ~/.udsdoc/.

CLI version

# ARXML in → LaTeX + compiled PDF out
udsdoc dcm.arxml cantp.arxml --dem dem.arxml --pdf -o spec.tex

# auto-discover a whole BSW repo (ARXML + C sources)
udsdoc --repo /path/to/bsw-project --pdf -o spec.tex

# DID / RID C sources for variable cross-referencing
udsdoc dcm.arxml --c-source did.c --c-source rid.c -o spec.tex

# Compliance Matrix + verify the concluded OEM↔supplier dialogue vs. the
# implementation, emit ISO 22901-1 PDX
udsdoc dcm.arxml --dem dem.arxml \
  --compliance-matrix matrix.csv --verify --verify-mini \
  --format pdx -o out/spec

Inputs — ARXML as positional arguments (+ --dem), or a whole repository via --repo; C sources via --c-source; requirements via --compliance-matrix (+ --compliance-matrix-sources). The matrix loader accepts the classic req_id/description/source_ref/traces_to contract and real-world OEM/DOORS exports (ID / ObjectType_<proj> / … / CONTENT / AcceptanceState_<proj> / CommentToCustomer_<proj> / Comment_Customer_<proj> / Final_Agreement_Status) — the per-project header tag (_ZFLSRM, …) is stripped automatically, heading/information rows are skipped, and Final_Agreement_Status is honored as the negotiation state (see the compliance_matrix module docstring for the full mapping).
Outputs — --format pdf / --pdf (Tex → PDF), --format pdx (ISO 22901-1 PDX; also tex, tex-zip — the structured multi-file LaTeX project as a ZIP — html, md, rst, odx). Exit codes: 0 ok / 1 runtime error / 2 usage error / 3 quality gate (--verify-strict, --validate, --odx-strict).

Python API:

from udsdoc import ArxmlParser, TexGenerator

spec = ArxmlParser().parse_multi(["dcm.arxml", "cantp.arxml"])
TexGenerator().generate(spec, "spec.tex")

Compliance verification

Verification is a local-first evidence pipeline:

--verify runs deterministic status, trace, numeric, access, and lexical checks with no network access.
--verify-mini adds the opt-in, dependency-free udsdoc-mini-semantic-v1 tier. It uses issue-scoped dialogue chronology, exact parsed-spec evidence, and mandatory abstention.
--verify-llm sends only the unresolved residue to the configured model.

Missing SID/DID/RID/DTC/session/security traces can be inferred from explicitly anchored English or Japanese text. Bare hexadecimal values are ignored, and every inferred trace retains confidence and source evidence. The frozen 90-row compact-model benchmark measures 64.4% coverage, 91.4% accuracy on covered rows, 80.0% deviation precision, and 40.0% deviation recall. The low recall is why uncertain remains a first-class review outcome.

The Verify feature (does the concluded OEM↔supplier dialogue match the implementation?) works with Anthropic (Claude), any OpenAI-compatible endpoint (OpenAI, Gemini's OpenAI-compatible API, OpenRouter, …) or local Ollama. Configure it once, whichever way suits you:

GUI — the LLM tab: sign in with Claude, paste a ChatGPT / Gemini key, or pick an Ollama model.
Env vars — UDSDOC_LLM_PROVIDER, UDSDOC_LLM_MODEL, UDSDOC_LLM_BASE_URL, UDSDOC_LLM_API_KEY (or a provider-native key such as ANTHROPIC_API_KEY / OPENAI_API_KEY / GEMINI_API_KEY).
CLI flags — --llm-provider, --llm-model, --llm-base-url, --llm-api-key, or --llm-config <file>.
Config file — ~/.udsdoc/llm.json ({"provider", "model", "base_url", "api_key", "timeout"}).

The static browser edition does not expose remote API-key providers. It can send uncertain rows to Ollama at the fixed loopback endpoint 127.0.0.1:11434, or run a consent-loaded WebLLM model in the browser. Model weights are not bundled. See the local-AI architecture and privacy contract for data-flow, offline, grounding, and reproducible benchmark details.

License

MIT

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

udsdoc-0.170.0.tar.gz (2.4 MB view details)

Uploaded Aug 3, 2026 Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

The dropdown lists show the available interpreters, ABIs, and platforms. Enable javascript to be able to filter the list of wheel files.

udsdoc-0.170.0-py3-none-any.whl (2.2 MB view details)

Uploaded Aug 3, 2026 Python 3

File details

Details for the file udsdoc-0.170.0.tar.gz.

File metadata

Download URL: udsdoc-0.170.0.tar.gz
Upload date: Aug 3, 2026
Size: 2.4 MB
Tags: Source
Uploaded using Trusted Publishing? Yes
Uploaded via: twine/7.0.0 CPython/3.13.14

File hashes

Hashes for udsdoc-0.170.0.tar.gz
Algorithm	Hash digest
SHA256	`23c07b8e6262610ab5488a5b4998ce36080fab2285877ddd59b30a0b03580c99`
MD5	`af29d800aa10216682217b5167a7068f`
BLAKE2b-256	`3db66e3241a3acb5fa315f9e03a9519ea2b6a850ac89d2d92010f6996a7991a8`

See more details on using hashes here.

File details

Details for the file udsdoc-0.170.0-py3-none-any.whl.

File metadata

Download URL: udsdoc-0.170.0-py3-none-any.whl
Upload date: Aug 3, 2026
Size: 2.2 MB
Tags: Python 3
Uploaded using Trusted Publishing? Yes
Uploaded via: twine/7.0.0 CPython/3.13.14

File hashes

Hashes for udsdoc-0.170.0-py3-none-any.whl
Algorithm	Hash digest
SHA256	`267340b7dbf706836003f53245cddebad9ad1ee2f67899ee7659cb7245320524`
MD5	`69d6382cdd48d44215eed720c99d90b7`
BLAKE2b-256	`058102cb3cecd052ad8f4c0754d8cf184fcf1083d7be75d87d32adee58ba920a`