This release adds 3 notable features for engineering teams evaluating rollout.
✓ No known CVEs patched in this version
Topics
ReleasePort's take
Light signalVersion v3.14.0 introduces LLM usage‑limit configuration (LLM_MAX_REQUESTS, LLM_MAX_TOKENS) and adds the get_person and get_family lookup tools.
Why it matters: The new config parameters let operators enforce strict request and token quotas; developers gain direct person/family lookups via Gramps IDs in AI chat flows.
Summary
AI summaryAdded LLM usage limits config and new lookup tools get_person and get_family.
Changes in this release
| Type | Severity | Summary | CVE |
|---|---|---|---|
| Feature | Medium |
Adds config parameters LLM_MAX_REQUESTS and LLM_MAX_TOKENS for usage limits Adds config parameters LLM_MAX_REQUESTS and LLM_MAX_TOKENS for usage limits Source: llm_adapter@2026-05-22 Confidence: high |
— |
| Feature | Medium |
Adds new tools get_person and get_family for Gramps ID lookups Adds new tools get_person and get_family for Gramps ID lookups Source: llm_adapter@2026-05-22 Confidence: high |
— |
| Feature | Medium |
Allows search tool filtering by object type Allows search tool filtering by object type Source: llm_adapter@2026-05-22 Confidence: high |
— |
| Feature | Medium |
Enables progress callback to report background task actions Enables progress callback to report background task actions Source: llm_adapter@2026-05-22 Confidence: high |
— |
| Feature | Medium |
Allows passing full message history, including tool calls and outputs, to the API Allows passing full message history, including tool calls and outputs, to the API Source: llm_adapter@2026-05-22 Confidence: high |
— |
| Feature | Medium |
Automatically trims message history when it grows too large Automatically trims message history when it grows too large Source: llm_adapter@2026-05-22 Confidence: high |
— |
| Feature | Medium |
Allows Markdown in API output Allows Markdown in API output Source: llm_adapter@2026-05-22 Confidence: high |
— |
| Feature | Medium |
All object endpoints now accept optional `handles` parameter to fetch multiple objects by handle All object endpoints now accept optional `handles` parameter to fetch multiple objects by handle Source: llm_adapter@2026-05-22 Confidence: high |
— |
| Feature | Medium |
Filters can now contain nested queries involving other object types without custom filter definition Filters can now contain nested queries involving other object types without custom filter definition Source: llm_adapter@2026-05-22 Confidence: high |
— |
Full changelog
AI Assistant Improvements
This release contains several improvements and fixes for the AI chat endpoint.
- Added more tailored and reliable usage limits: new config parameters
LLM_MAX_REQUESTSandLLM_MAX_TOKENS. - New tools
get_personandget_familyto look up individual people or families by Gramps ID - Search tool: allow filtering by object type
- Progress callback: background task now report what they are doing while the agent is working (which tools is being called)
- Allow passing the full message history, including tool calls and outputs, to the API
- Automatically trim the history when it grows too large
- Allow Markdown in the output (matching an upcoming frontend improvement)
Other Improvements
- All object endpoints now accept an optional
handlesparameter that allows to fetch more than one object at a time by passing an explicit handle - Filters can now contain nested involving other object types, e.g. filtering for people that have events matching a specific rule, without creating a named custom filter first
Full Changelog: https://github.com/gramps-project/gramps-web-api/compare/v3.13.1...v3.14.0
Weekly OSS security release digest.
The CVE patches and breaking changes that affected production tools this week. One email, every Sunday.
No spam, unsubscribe anytime.
Share this release
About Gramps Web
Web app for collaborative genealogy, based on and interoperable with Gramps, the open source genealogy desktop application.
Beta — feedback welcome: [email protected]