chore(eco-watch): 2026-04-18 survey — AMD GAIA + ClawRun - #369
Merged
Merged
Conversation
Add two new entries to docs/ecosystem-watch.md: - **AMD GAIA** (amd/gaia, ~1.2k ⭐, MIT, v0.17.2 April 10 2026): AMD-backed local-first agent framework with MCP client support, RAG, vision, and voice. Hardware-locked to Ryzen AI but signals local/privacy-first positioning. @tool decorator pattern worth borrowing for workspace adapters. - **ClawRun** (clawrun-sh/clawrun, ~84 ⭐, Apache 2.0, 45 releases): Closest architectural match we've tracked — hosting/lifecycle layer with sandbox, heartbeat, snapshot/resume, channels, and cost tracking. Per-channel budget enforcement is a concrete gap in our workspace_channels. Filed #368. HEAD at survey time: a4a89a3 Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
molecule-ai Bot
pushed a commit
that referenced
this pull request
Apr 21, 2026
All CI green. Docs-only: adds AMD GAIA + ClawRun ecosystem survey entries.
HongmingWang-Rabbit
pushed a commit
that referenced
this pull request
Jun 12, 2026
… na-gate fallback PRs with thousands of bot-relay comments (e.g. mc#1242-class history) pushed `get_issue_comments` past the runner's cgroup memory cap and 137'd the gate job. Sub-agent aa8fa266 verified via SHA64 1:1 mapping that the leak source is `.gitea/scripts/sop-checklist.py:463-484` — a `while True` pagination that accumulates the FULL Gitea-API comment dict (~2 KiB median, ~3 KiB p95) into one list, then iterates it twice (compute_ack_state + compute_na_state). Fix shape (task #369): 1. `get_issue_comments` now projects to a minimal `{user.login, body}` shape during pagination — drops html_url, pull_request_url, assets, created_at/updated_at, user.avatar_url, etc. Synthetic benchmark on a 5000-comment fixture: 8.2 MiB → 5.5 MiB peak heap (~33% smaller). 2. New `iter_issue_comments` generator yields one minimal-dict per comment, allowing future streaming consumers. 3. Per-comment body cap (8 KiB, env-tunable) — the directive parser only walks for ^/sop-* markers in the first few KiB; a single pasted-CI-log comment can no longer OOM the runner. 4. Script-entry `RLIMIT_AS = 2 GiB` (skipped under SOP_CHECKLIST_NO_RLIMIT=1 for the test suite) so any future regression surfaces as a catchable MemoryError instead of SIGKILL. 5. `--max-comments` cap (default 5000, env-tunable) — at the cap we post a SOFT pending status with a `[volume-skipped]` note so neither BP nor the author gets stuck. No-block. 6. Fold in `probe()` n/a-gate fallback (was issue-355-class KeyError 'security-review'): when called from `compute_na_state` with a gate name that isn't also an item slug, fall back to the gate's own `required_teams` from config instead of KeyError'ing on `items_by_slug[slug]`. Tests: 88/88 pass (87 existing + 8 new for streaming + body-cap + na-gate fallback). Live dry-run against open PR #1608 succeeds end- to-end with state=failure (expected, no acks yet) — minimal-dict shape flows cleanly through compute_ack_state + render_status. Closes #369 follow-up to #364 OOM source. RCA: sub-agent aa8fa266 (SHA64 1:1 mapping, no overlap-correlation). Supersedes mis-targeted attempts #365 (qa-review/security-review bash) and #366 (workflow yaml). Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Daily eco-watch survey for 2026-04-18. Two new entries added to
docs/ecosystem-watch.md; one GH issue filed.AMD GAIA (
amd/gaia, ~1.2k ⭐, MIT, v0.17.2 April 10 2026) — AMD-backed local-first agent framework. MCP client support, RAG, vision, voice. Hardware-locked to Ryzen AI but represents a credible "local, privacy-first, no cloud dependency" positioning signal.@tooldecorator pattern is worth borrowing. No issue filed — no direct gap surfaced.ClawRun (
clawrun-sh/clawrun, ~84 ⭐, Apache 2.0, 45 releases) — Closest architectural match to Molecule AI tracked to date. Hosting + lifecycle layer: sandbox, heartbeat, snapshot/resume, channels (Telegram/Discord/Slack/WhatsApp), cost tracking, budget enforcement. Per-channel budget enforcement is a concrete gap in ourworkspace_channels. → Filed feat(channels): per-channel budget enforcement (inspired by ClawRun) #368: addbudget_limit_usdper channel.Skipped today (already tracked): hermes-agent, GenericAgent, AgentScope, OpenSRE, claude-mem, open-agents, plannotator, Archon, claude-code-routines (all verified against trending).
HN/web signal: AMD GAIA HN item #47756772. No other new framework launches today.
HEAD at survey time:
a4a89a3Test plan
###entries appear under## Entries, above## Candidates to add🤖 Generated with Claude Code