Pi Web Access Files

Pi Web Access 0.27.0 gives fetch_content users more control over page fetching and answer generation. You can now tune direct HTTP and Jina Reader timeouts from config, and answer mode can use a default model without repeating it on each call. The release also makes HTML fallback parsing quieter on pages with relative canonical links. GitHub repository fetching is safer when multiple Pi processes run at the same time.

Highlights:

• Configure direct HTTP and Jina Reader fetch timeouts in web-search.json.
• Set default answer-mode provider and model while keeping per-call overrides.
• Avoid noisy Defuddle fallback warnings on relative canonical URLs.
• Keep GitHub clone cleanup isolated across concurrent Pi runtimes.

Full changelog:

Highlights

• fetch_content now lets you tune direct HTTP and Jina Reader timeouts from config.
• Answer mode can use a configured default model while still allowing per-call overrides.
• HTML fallback parsing is quieter for pages with relative canonical links.
• GitHub repository fetching is safer when more than one Pi process is running.

Added

• Added opt-in fetch.timeout configuration in seconds for the direct HTTP and Jina Reader fetch_content paths, with per-call timeout overrides taking precedence. Thanks to @linuxtextadventurer for PR [#327].
• Added opt-in fetch.answerProvider and fetch.answerModel defaults for fetch_content answer mode, with per-call answerModel overrides taking precedence. Thanks to @linuxtextadventurer for PR [#328].

Fixed

• Set the Defuddle fallback document URL before parsing pages with relative canonical links, preventing ERR_INVALID_URL warnings. Thanks to @bin115885 for issue [#322].
• Isolated GitHub clone workdirs per extension runtime so cleanup in one process cannot delete another process's clone. Thanks to @MDGChamomile for PR [#323].