Skip to content

RepoDNA 1.1.0

Choose a tag to compare

View all tags
@sanskarIN sanskarIN released this 04 Oct 05:08
· 406 commits to main since this release
v1.1.0
a168641
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

Optional AI explanations: prose about a repository, grounded in the analysis, checked
against its evidence, and off until you configure a provider.

Explanations

  • repodna explain explains a repository (--about repository, architecture,
    history, dependencies, or onboarding), one module (--module), one hotspot
    (--hotspot), or answers a question (--ask), as text, Markdown, or JSON with its
    provenance: provider, model, revision, cited evidence, and token use.
  • The model receives a numbered selection of evidence from the analysis, never the
    repository, and its answer is checked against it: citations of evidence that was not
    sent are removed, statements without a valid citation are labeled as not supported, and
    statements the model marks as inferences are labeled as such.
  • Three providers: a local program that reads the prompt on standard input (for example
    ollama run <model>), any OpenAI-compatible server (Ollama, llama.cpp, LM Studio, vLLM,
    or a hosted service), and the Anthropic API.
  • --dry-run shows the exact prompt, the evidence, where it would be sent, and the
    highest cost from prices you set, without contacting anything.
  • Answers are cached on this machine; --fresh asks again and repodna cache clear
    removes them. When a model declines, RepoDNA reports the reason the provider gives.
  • repodna doctor reports the configured provider without contacting it.

Privacy and safety

  • AI is set only in your user configuration: a repository's own configuration cannot
    enable it or allow remote providers.
  • A provider whose endpoint is not on this machine is refused unless you allow remote AI
    with privacy.remote_ai = true or, for one run, --allow-remote-ai.
  • API keys are read from an environment variable you name and never stored. Without one,
    ANTHROPIC_API_KEY is sent only to Anthropic's own endpoint.
  • No source code is sent unless you enable ai.include_source_excerpts: excerpts are a
    few lines per file, have likely secrets redacted, and are never read through a link
    that leads out of the checkout. Contributor names and email addresses are never sent.
  • The model gets no tools: it cannot run commands, read files, or use the network through
    RepoDNA. Repository text is sent as data, and answers are cleaned of terminal control
    sequences and escaped in Markdown.
  • RepoDNA uses the network only to clone a Git URL you give and to reach an AI provider
    you configure.

Web interface and desktop app

  • The start page, the settings, and the analysis details say that explanations are
    optional and need a provider you configure, and whether an analysis used remote AI.

Downloads

  • repodna-1.1.0-<target>: the repodna command line, with the web interface of
    repodna serve built in. The Linux builds are static and run on any distribution.
  • RepoDNA_1.1.0_amd64.deb, RepoDNA-1.1.0-1.x86_64.rpm,
    RepoDNA_1.1.0_universal.dmg, RepoDNA_1.1.0_x64_en-US.msi, and
    RepoDNA_1.1.0_x64-setup.exe: the desktop app.
  • ghcr.io/sanskarin/repodna:1.1.0: a container image with the command line and
    Git, for CI jobs: docker run --rm -v "$PWD:/work" ghcr.io/sanskarin/repodna analyze .
  • SHA256SUMS.txt: checksums of every file.

The binaries are not code-signed, so macOS and Windows ask for confirmation the first time they start; see the installation guide.

Assets 13
Loading