Navigara

github.com-google-adk-python

all · 18 devs · built 2026-06-13

Repository snapshot

Performance · 90d

32.1ETV

+27.0% since Q2 2025

ETV / dev / mo

1.2ETV

10 devs · 30-day window

Last 7 days · per dev

0.7ETV

per developer per week

Work mix

Growth31%· 55.0Maintenance53%· 93.8Fixes17%· 29.4
178.1 ETV all-time1,565 commits18 all-time devsApr 2025 – Jun 2026

Monthly reports

  • Highlights

    • Introduced new *AI agent skills* including `adk-pr-triage` for automated pull request management and CLA verification [ce9011c1 · Bo Yang], `adk-review` for rigorous change quality control [cc6f78c3 · Bo Yang], and `adk-issue` for GitHub issue analysis and triage [be03166f · Bo Yang].
    • Achieved a *performance optimization* in *LLM flows* by enabling parallel resolution of agent tool unions, significantly reducing execution time for agent operations [ae95a972 · Google Team Member].
    • Enhanced *observability* by adding native OpenTelemetry `gen_ai.client.*` metrics with `gen_ai.agent.name` attribute for more granular insights into LLM operations [0bb329ba · Marcin Vinietowicz].
    • Added a new `RubricBasedMultiTurnTrajectoryEvaluator` to the *evaluation subsystem*, enabling sophisticated LLM-based assessment of multi-turn conversational quality [cae23371 · wenhzhang].
    • Implemented critical bug fixes to prevent *MCP tool* initialization hangs and task group leaks by switching to `anyio.fail_after` for timeouts [334ef815 · Bo Yang].
    • Addressed a security vulnerability in the `ReadFileTool` by properly escaping `path` and `range` arguments to prevent *shell injection* [e16629b3 · Wei (Jack) Sun].
    • Performed significant *refactoring* to restore backward compatibility for 1.x agent configurations and split *environment tools* into single-class modules, improving maintainability and simplifying configuration [44cd1167 · George Weale], [1cc298ed · Wei (Jack) Sun].

    Observations

    • The *Grow* score for May 2026 was 4, representing a substantial 40% decrease compared to the 5-month average of 7, indicating a reduced focus on new feature development during this period.
    • Total output (Grow + Maintenance + Waste) decreased by 21% (18 current vs 23 5-month average), despite the commit volume remaining stable at 169 commits, which is only +1% above the 5-month average of 167 commits.
    • A notable number of commits focused on *bug fixes* and *stability improvements* across various components, including *workflow rehydration* [4f992b0c · Samarth2410], *session management* [03ef3f61 · George Weale], *Gemini LLM connection* [e896c620 · Kathy Wu], and *Firestore session integration* [a5db3467 · Kjell Reidar Jøssang], suggesting an emphasis on hardening existing systems.
    • Several *maintenance* activities were undertaken, such as refactoring agent and tool retrieval functions for *asynchronous execution* [e623b3b4 · Bo Yang], graduating the `SkillToolset` from experimental status [91ded3a1 · Kathy Wu], and updating CI/CD workflows for security checks and test timeouts [eaff9c02 · Google Team Member], [7ad79947 · asobran].
    • The `tools` subsystem, particularly `AgentTool` and `SkillToolset`, received multiple fixes and enhancements, addressing issues like plugin lifecycle management [2a68c4e7 · thacht], skill script import paths [9296198b · Yufeng He], and `LoadSkillResourceTool` retry loops [bc45ee67 · Google Team Member], indicating an area of active refinement and complexity.

Performance over time

ETV stacked by Growth, Maintenance and Fixes — 90-day moving average, normalized to ETV / month.

Average performance per developer

ETV per active developer per month — 30-day moving average.

Active developers over time

Unique developers committing each day — 90-day moving average.

Knowledge concentration

How dependent is this repo on a small number of contributors? Higher top-1 share = higher key-person risk.

Top 1
21.5 %
Top 3
48.6 %
Top 5
67.2 %

Xiang (Sean) Zhou owns 21.5 % of commits.

Top contributors

Most impactful commits

Top 20 by ETV in the all-time window.