Spaces:
Configuration error
Configuration error
| <html lang="en"> | |
| <head> | |
| <meta charset="utf-8"> | |
| <meta name="viewport" content="width=device-width, initial-scale=1"> | |
| <title>Localpager GEPA Final Report</title> | |
| <style> | |
| :root { --bg:#f6f8fb; --panel:#fff; --ink:#172033; --muted:#667085; --line:#d8dee9; --blue:#185abc; --green:#137333; --red:#b42318; --amber:#ad5f00; } | |
| * { box-sizing:border-box; } | |
| body { margin:0; background:var(--bg); color:var(--ink); font-family:-apple-system,BlinkMacSystemFont,"Segoe UI",Roboto,Arial,sans-serif; line-height:1.45; } | |
| header { background:#fff; border-bottom:1px solid var(--line); padding:28px 32px 20px; position:sticky; top:0; z-index:2; } | |
| h1 { margin:0 0 8px; font-size:28px; } | |
| main { max-width:1280px; margin:0 auto; padding:24px 32px 52px; } | |
| nav { display:flex; gap:10px; flex-wrap:wrap; margin-top:14px; } | |
| a { color:#0b57d0; text-decoration:none; } | |
| nav a, .button { border:1px solid #c9dcff; background:#eef4ff; border-radius:7px; padding:7px 10px; } | |
| section { background:var(--panel); border:1px solid var(--line); border-radius:8px; padding:20px; margin-bottom:20px; } | |
| h2 { margin:0 0 14px; font-size:20px; } | |
| h3 { margin:18px 0 10px; font-size:16px; } | |
| .gridcards { display:grid; grid-template-columns:repeat(auto-fit,minmax(180px,1fr)); gap:12px; } | |
| .card { border:1px solid var(--line); border-radius:8px; padding:14px; background:#fbfcfe; min-height:104px; } | |
| .card .label { color:var(--muted); font-size:12px; text-transform:uppercase; letter-spacing:.04em; } | |
| .card .value { margin-top:8px; font-size:26px; font-weight:750; } | |
| .card .sub { margin-top:5px; color:var(--muted); font-size:13px; } | |
| .good { color:var(--green); font-weight:700; } .bad { color:var(--red); font-weight:700; } .neutral,.muted { color:var(--muted); } | |
| .scroll { overflow:auto; border:1px solid var(--line); border-radius:8px; } | |
| table { width:100%; border-collapse:collapse; font-size:13px; } | |
| th,td { border-bottom:1px solid var(--line); padding:8px 9px; text-align:left; vertical-align:top; } | |
| th { background:#f0f3f8; color:#384152; font-weight:700; line-height:1.2; white-space:nowrap; vertical-align:middle; } | |
| tr.selected { background:#eefbf2; } | |
| code { font-family:ui-monospace,SFMono-Regular,Consolas,monospace; font-size:12px; } | |
| .chart-wrap { border:1px solid var(--line); border-radius:8px; background:#fff; overflow-x:auto; } | |
| svg { min-width:900px; width:100%; height:auto; display:block; } | |
| .grid { stroke:#e5eaf2; } .tick { fill:#667085; font-size:12px; } .label { font-size:11px; } .point { fill:#1f2937; font-size:12px; font-weight:700; } .axis-label { fill:#334155; font-size:13px; } | |
| .note { border-left:4px solid var(--amber); padding:10px 12px; background:#fff8eb; color:#4a2d00; } | |
| @media (max-width:800px) { header { position:static; } main,header { padding-left:18px; padding-right:18px; } } | |
| </style> | |
| </head> | |
| <body> | |
| <header> | |
| <h1>Localpager GEPA Final Report</h1> | |
| <div class="muted">Updated 2026-06-14T13:35:45Z. Dataset: Shaun ordered 60-row set. Gold labels: canonical <code>ds4.jsonl</code>. Task model: local 12B. Concurrency: 2.</div> | |
| <nav><a href="#summary">Summary</a><a href="#trajectory">Score Trajectory</a><a href="#metrics">Metrics</a><a href="#gates">Gates</a><a href="#misses">Remaining Misses</a><a href="gepa-12b-row30-prop20-continuation-20260614T021448Z/score_report.html">GEPA Iteration Graph</a><a href="prompt-diffs/index.html">Prompt Diffs</a></nav> | |
| </header> | |
| <main> | |
| <section id="summary"> | |
| <h2>Promoted Result</h2> | |
| <div class="gridcards"><div class="card"><div class="label">Final mean score</div><div class="value good">0.6233</div><div class="sub">delta 0.1911 vs v9.1</div></div><div class="card"><div class="label">Micro-F1</div><div class="value good">0.7985</div><div class="sub">target >= 0.7000</div></div><div class="card"><div class="label">Precision / recall</div><div class="value good">0.9375 / 0.6954</div><div class="sub">FP weighted harder than FN</div></div><div class="card"><div class="label">False positives</div><div class="value good">7</div><div class="sub">target <= 20</div></div><div class="card"><div class="label">False negatives</div><div class="value good">46</div><div class="sub">target <= 58</div></div><div class="card"><div class="label">Over-label events</div><div class="value good">0</div><div class="sub">target 0 or 1</div></div><div class="card"><div class="label">Structural failures</div><div class="value good">0</div><div class="sub">target 0</div></div><div class="card"><div class="label">Mean predicted labels</div><div class="value good">1.8667</div><div class="sub">same as v9.1 baseline</div></div></div> | |
| <p class="note">The promoted v3 candidate clears the strict gates and keeps mean predicted labels at the v9.1 baseline. The main caveat is scientific, not mechanical: the manual repair used mistakes from this same 60-row set, so a fresh holdout is still needed before treating it as deployment evidence.</p> | |
| </section> | |
| <section id="trajectory"> | |
| <h2>60-Row Score Trajectory</h2> | |
| <div class="chart-wrap"><svg viewBox="0 0 920 310" role="img" aria-label="60-row score trajectory"> | |
| <rect width="100%" height="100%" fill="#fff"/> | |
| <line x1="64" y1="244.0" x2="892" y2="244.0" class="grid"/> | |
| <text x="54" y="248.0" text-anchor="end" class="tick">0.38</text> | |
| <line x1="64" y1="189.0" x2="892" y2="189.0" class="grid"/> | |
| <text x="54" y="193.0" text-anchor="end" class="tick">0.48</text> | |
| <line x1="64" y1="134.0" x2="892" y2="134.0" class="grid"/> | |
| <text x="54" y="138.0" text-anchor="end" class="tick">0.59</text> | |
| <line x1="64" y1="79.0" x2="892" y2="79.0" class="grid"/> | |
| <text x="54" y="83.0" text-anchor="end" class="tick">0.70</text> | |
| <line x1="64" y1="24.0" x2="892" y2="24.0" class="grid"/> | |
| <text x="54" y="28.0" text-anchor="end" class="tick">0.80</text> | |
| <polyline points="64.0,216.6 229.6,185.8 395.2,162.5 560.8,132.1 726.4,38.2 892.0,116.5" fill="none" stroke="#185abc" stroke-width="3" stroke-linecap="round" stroke-linejoin="round"/> | |
| <circle cx="64.0" cy="216.6" r="6" fill="#185abc" stroke="#fff" stroke-width="2"><title>v9.1 seed 0.4322</title></circle> | |
| <text x="64.0" y="204.6" text-anchor="middle" class="point">0.4322</text> | |
| <text x="64.0" y="274" text-anchor="middle" class="tick label">v9.1 seed</text> | |
| <circle cx="229.6" cy="185.8" r="6" fill="#185abc" stroke="#fff" stroke-width="2"><title>GEPA six 0.4912</title></circle> | |
| <text x="229.6" y="173.8" text-anchor="middle" class="point">0.4912</text> | |
| <text x="229.6" y="274" text-anchor="middle" class="tick label">GEPA six</text> | |
| <circle cx="395.2" cy="162.5" r="6" fill="#185abc" stroke="#fff" stroke-width="2"><title>Previous proper best 0.5355</title></circle> | |
| <text x="395.2" y="150.5" text-anchor="middle" class="point">0.5355</text> | |
| <text x="395.2" y="274" text-anchor="middle" class="tick label">Previous proper best</text> | |
| <circle cx="560.8" cy="132.1" r="6" fill="#185abc" stroke="#fff" stroke-width="2"><title>Prop20 best 0.5936</title></circle> | |
| <text x="560.8" y="120.1" text-anchor="middle" class="point">0.5936</text> | |
| <text x="560.8" y="274" text-anchor="middle" class="tick label">Prop20 best</text> | |
| <circle cx="726.4" cy="38.2" r="6" fill="#185abc" stroke="#fff" stroke-width="2"><title>Hardcase repair v2 0.7729</title></circle> | |
| <text x="726.4" y="26.2" text-anchor="middle" class="point">0.7729</text> | |
| <text x="726.4" y="274" text-anchor="middle" class="tick label">Hardcase repair v2</text> | |
| <circle cx="892.0" cy="116.5" r="6" fill="#15803d" stroke="#fff" stroke-width="2"><title>Cardinality repair v3 0.6233</title></circle> | |
| <text x="892.0" y="104.5" text-anchor="middle" class="point">0.6233</text> | |
| <text x="892.0" y="274" text-anchor="middle" class="tick label">Cardinality repair v3</text> | |
| <text x="478.0" y="300" text-anchor="middle" class="axis-label">candidate timeline</text> | |
| </svg></div> | |
| </section> | |
| <section id="metrics"> | |
| <h2>Candidate Metrics</h2> | |
| <div class="scroll"><table><thead><tr><th>Candidate</th><th>Mean</th><th>Delta</th><th>Precision</th><th>Recall</th><th>F1</th><th>FP</th><th>FN</th><th>Over</th><th>Struct</th><th>Exact</th><th>Mean Labels</th><th>Label Delta</th><th>Source</th></tr></thead><tbody><tr class=''><td><b>v9.1 seed</b><br><span class='muted'>baseline</span></td><td>0.4322</td><td class='neutral'>0.0000</td><td>0.7411</td><td>0.5804</td><td>0.6510</td><td>29</td><td>60</td><td>2</td><td>3</td><td>10</td><td>1.8667</td><td>0.0000</td><td><a href='dashboard-20260613-gepa/summary.json'>source</a></td></tr><tr class=''><td><b>GEPA six</b><br><span class='muted'>early GEPA</span></td><td>0.4912</td><td class='good'>0.0590</td><td>0.7965</td><td>0.5960</td><td>0.6818</td><td>23</td><td>61</td><td>1</td><td>0</td><td>12</td><td>n/a</td><td></td><td><a href='dashboard-20260613-gepa/summary.json'>source</a></td></tr><tr class=''><td><b>Previous proper best</b><br><span class='muted'>proper GEPA</span></td><td>0.5355</td><td class='good'>0.1033</td><td>0.7778</td><td>0.6490</td><td>0.7076</td><td>28</td><td>53</td><td>5</td><td>2</td><td>21</td><td>2.1000</td><td>0.2333</td><td><a href='validation-12b-proper-best-20260613T111155Z/gepa-12b-proper-best-limit60.json'>source</a></td></tr><tr class=''><td><b>Prop20 best</b><br><span class='muted'>proper GEPA</span></td><td>0.5936</td><td class='good'>0.1613</td><td>0.8707</td><td>0.6689</td><td>0.7566</td><td>15</td><td>50</td><td>2</td><td>4</td><td>24</td><td>1.9333</td><td>0.0667</td><td><a href='validation-12b-row30-prop16-best-20260614T081931Z/gepa-12b-row30-prop16-best-limit60.json'>source</a></td></tr><tr class=''><td><b>Hardcase repair v2</b><br><span class='muted'>manual repair, rejected</span></td><td>0.7729</td><td class='good'>0.3407</td><td>0.9343</td><td>0.8477</td><td>0.8889</td><td>9</td><td>23</td><td>4</td><td>0</td><td>38</td><td>2.2833</td><td>0.4167</td><td><a href='validation-manual-hardcase-repair-v2-max4096-20260614T100359Z/manual-hardcase-repair-v2-max4096-limit60.json'>source</a></td></tr><tr class='selected'><td><b>Cardinality repair v3</b><br><span class='muted'>promoted</span></td><td>0.6233</td><td class='good'>0.1911</td><td>0.9375</td><td>0.6954</td><td>0.7985</td><td>7</td><td>46</td><td>0</td><td>0</td><td>19</td><td>1.8667</td><td>0.0000</td><td><a href='validation-manual-hardcase-repair-v3-cardinality-max4096-20260614T103814Z/manual-hardcase-repair-v3-cardinality-max4096-limit60.json'>source</a></td></tr></tbody></table></div> | |
| </section> | |
| <section id="gates"> | |
| <h2>Strict Gate Check</h2> | |
| <div class="scroll"><table><thead><tr><th>Gate</th><th>Target</th><th>Observed</th><th>Result</th></tr></thead><tbody><tr><td>mean weighted score</td><td>>= 0.5400</td><td>0.6233</td><td class='good'>pass</td></tr><tr><td>score delta vs v9.1</td><td>>= +0.1000</td><td>0.1911</td><td class='good'>pass</td></tr><tr><td>micro-F1</td><td>>= 0.7000</td><td>0.7985</td><td class='good'>pass</td></tr><tr><td>precision</td><td>>= 0.8000</td><td>0.9375</td><td class='good'>pass</td></tr><tr><td>recall</td><td>>= 0.6100</td><td>0.6954</td><td class='good'>pass</td></tr><tr><td>false positives</td><td><= 20</td><td>7</td><td class='good'>pass</td></tr><tr><td>false negatives</td><td><= 58</td><td>46</td><td class='good'>pass</td></tr><tr><td>over-label events</td><td>0 or 1</td><td>0</td><td class='good'>pass</td></tr><tr><td>structural failures</td><td>0</td><td>0</td><td class='good'>pass</td></tr><tr><td>exact matches</td><td>>= 15</td><td>19</td><td class='good'>pass</td></tr><tr><td>mean predicted-label delta</td><td><= +0.10</td><td>0.0000</td><td class='good'>pass</td></tr></tbody></table></div> | |
| </section> | |
| <section id="links"> | |
| <h2>Updated Links</h2> | |
| <p><a class="button" href="gepa-12b-row30-prop20-continuation-20260614T021448Z/score_report.html">Prop20 GEPA score report</a> <a class="button" href="gepa-12b-row30-prop20-continuation-20260614T021448Z/candidate_tree.html">Prop20 candidate tree</a> <a class="button" href="prompt-diffs/index.html">Prompt diffs</a></p> | |
| <p class="muted">Promoted routing-policy SHA-256: <code>b2576ca027148e109a1e72029c192ea9e26be486508671ec7c11025ac80f948b</code>.</p> | |
| </section> | |
| <section id="misses"> | |
| <h2>Remaining v3 Misses</h2> | |
| <p class="muted">Rows where v3 is not an exact match. These are the next useful targets for a clean holdout-aware follow-up.</p> | |
| <div class="scroll"><table><thead><tr><th>Target</th><th>Title</th><th>Score</th><th>Gold</th><th>Predicted</th><th>FP</th><th>FN</th></tr></thead><tbody><tr><td><a href='https://github.com/openclaw/openclaw/pull/44379'>#44379</a></td><td>fix(pi-runner): harden context-overflow recovery with one suppress-hook retry</td><td>0.1667</td><td>coding_agents, memory, hooks, reliability</td><td>agent_runtime, reliability</td><td>agent_runtime</td><td>coding_agents, memory, hooks</td></tr><tr><td><a href='https://github.com/openclaw/openclaw/pull/45393'>#45393</a></td><td>fix(errors): friendly message and last-message repair for tool_use/tool_result mismatch (#45385)</td><td>0.2000</td><td>tool_calling, coding_agents, reliability</td><td>tool_calling, agent_runtime</td><td>agent_runtime</td><td>coding_agents, reliability</td></tr><tr><td><a href='https://github.com/openclaw/openclaw/pull/47083'>#47083</a></td><td>fix: respect totalTokensFresh flag to avoid showing stale token counts</td><td>0.2000</td><td>sessions, telemetry_usage</td><td>ui_tui</td><td>ui_tui</td><td>sessions, telemetry_usage</td></tr><tr><td><a href='https://github.com/openclaw/openclaw/pull/81957'>#81957</a></td><td>ci: harden GitHub Actions supply-chain boundaries</td><td>0.2500</td><td>security</td><td>packaging_deployment</td><td>packaging_deployment</td><td>security</td></tr><tr><td><a href='https://github.com/openclaw/openclaw/pull/65364'>#65364</a></td><td>feat(plugins): add registerProviderRuntimeAuthOverride API</td><td>0.2500</td><td>auth_identity, api_surface</td><td>skills_plugins, auth_identity</td><td>skills_plugins</td><td>api_surface</td></tr><tr><td><a href='https://github.com/openclaw/openclaw/pull/80008'>#80008</a></td><td>feat(plugins): expose ACP spawn and prompt in plugin runtime</td><td>0.2500</td><td>acp, coding_agents</td><td>skills_plugins, acp</td><td>skills_plugins</td><td>coding_agents</td></tr><tr><td><a href='https://github.com/openclaw/openclaw/issues/90146'>#90146</a></td><td>google-vertex: Missing gemini-3.1-flash-lite in provider catalog causes silent failure instead of error</td><td>0.2500</td><td>local_model_providers, reliability</td><td>local_model_providers, model_serving</td><td>model_serving</td><td>reliability</td></tr><tr><td><a href='https://github.com/openclaw/openclaw/issues/73910'>#73910</a></td><td>BUG: OpenClaw-managed Codex ACP uses isolated CODEX_HOME without auth bridge and sends unsupported timeout config</td><td>0.3333</td><td>codex, acp, acpx, auth_identity</td><td>codex, acp</td><td>none</td><td>acpx, auth_identity</td></tr><tr><td><a href='https://github.com/openclaw/openclaw/issues/52249'>#52249</a></td><td>ACP parent session stuck until refresh when yielded waiting for child completion</td><td>0.5000</td><td>acp, sessions, reliability</td><td>acp, sessions</td><td>none</td><td>reliability</td></tr><tr><td><a href='https://github.com/openclaw/openclaw/issues/83863'>#83863</a></td><td>ACP/Codex child tasks can be marked succeeded with progress-only output and no final deliverable</td><td>0.5000</td><td>acp, codex, agent_runtime</td><td>acp, codex</td><td>none</td><td>agent_runtime</td></tr><tr><td><a href='https://github.com/openclaw/openclaw/pull/48940'>#48940</a></td><td>ACP: add gateway-owned node-backed runtime</td><td>0.5000</td><td>acp, gateway, agent_runtime</td><td>acp, gateway</td><td>none</td><td>agent_runtime</td></tr><tr><td><a href='https://github.com/openclaw/openclaw/issues/48580'>#48580</a></td><td>Bug: acpx codex sessions 创建的会话立即退出 - stdin is not a terminal</td><td>0.5000</td><td>acpx, codex, sessions</td><td>acpx, codex</td><td>none</td><td>sessions</td></tr><tr><td><a href='https://github.com/openclaw/openclaw/issues/39248'>#39248</a></td><td>Bug: sandbox.mode: "non-main" silently breaks sessions_spawn subagent initialization</td><td>0.5000</td><td>coding_agents, sandboxing, agent_runtime</td><td>sandboxing, agent_runtime</td><td>none</td><td>coding_agents</td></tr><tr><td><a href='https://github.com/openclaw/openclaw/issues/71216'>#71216</a></td><td>Config schema: add `sandbox`, `routing.rules`, `instances`, and `gateway.nodes.denyPaths`</td><td>0.5000</td><td>config, sandboxing, gateway</td><td>config, gateway</td><td>none</td><td>sandboxing</td></tr><tr><td><a href='https://github.com/openclaw/openclaw/issues/84477'>#84477</a></td><td>Discord embedded-run prep wedge before strict-agentic, recovery skips sessionId=unknown lanes</td><td>0.5000</td><td>sessions, agent_runtime, reliability</td><td>sessions, reliability</td><td>none</td><td>agent_runtime</td></tr><tr><td><a href='https://github.com/openclaw/openclaw/issues/51667'>#51667</a></td><td>Feature: Native Audio Input for Omni-Modal Models (skip STT transcription)</td><td>0.5000</td><td>model_serving, security, config</td><td>model_serving, config</td><td>none</td><td>security</td></tr><tr><td><a href='https://github.com/openclaw/openclaw/pull/43765'>#43765</a></td><td>Improve runtime recovery for heartbeat, Feishu, and exec sessions</td><td>0.5000</td><td>reliability, exec_tools, cron_automation</td><td>reliability, exec_tools</td><td>none</td><td>cron_automation</td></tr><tr><td><a href='https://github.com/openclaw/openclaw/pull/80783'>#80783</a></td><td>Policy: add model, network, and MCP conformance checks</td><td>0.5000</td><td>mcp_tooling, config, security</td><td>config, mcp_tooling</td><td>none</td><td>security</td></tr><tr><td><a href='https://github.com/openclaw/openclaw/issues/68187'>#68187</a></td><td>SSE-backed MCP sessions can stay stale after server restart and fail with 'Session not found'</td><td>0.5000</td><td>mcp_tooling, sessions, gateway</td><td>mcp_tooling, sessions</td><td>none</td><td>gateway</td></tr><tr><td><a href='https://github.com/openclaw/openclaw/issues/78528'>#78528</a></td><td>Security: skill SecretRef API keys still leak into exec child environments</td><td>0.5000</td><td>security, exec_tools, skills_plugins</td><td>security, exec_tools</td><td>none</td><td>skills_plugins</td></tr><tr><td><a href='https://github.com/openclaw/openclaw/issues/84715'>#84715</a></td><td>[Bug]: @openclaw/codex peer link failure reproduced on 2026.5.19 after update</td><td>0.5000</td><td>codex, packaging_deployment</td><td>codex</td><td>none</td><td>packaging_deployment</td></tr><tr><td><a href='https://github.com/openclaw/openclaw/issues/70529'>#70529</a></td><td>[Bug]: Desktop cannot use existing Chrome sessions: EasyClaw Google sign-in fails, and user profile attach fails with spawn npx ENOENT</td><td>0.5000</td><td>browser_automation, packaging_deployment</td><td>browser_automation</td><td>none</td><td>packaging_deployment</td></tr><tr><td><a href='https://github.com/openclaw/openclaw/issues/84757'>#84757</a></td><td>[Bug]: Telegram session can get stuck after compaction when encrypted reasoning content fails verification</td><td>0.5000</td><td>sessions, chat_integrations, reliability</td><td>sessions, chat_integrations</td><td>none</td><td>reliability</td></tr><tr><td><a href='https://github.com/openclaw/openclaw/issues/44202'>#44202</a></td><td>[Bug]: local memory embeddings on Apple Silicon can crash gateway in ggml-metal / node-llama-cpp; need official Metal/GPU guidance</td><td>0.5000</td><td>local_models, memory, self_hosted_inference</td><td>local_models, memory</td><td>none</td><td>self_hosted_inference</td></tr><tr><td><a href='https://github.com/openclaw/openclaw/issues/10467'>#10467</a></td><td>[Feature Request]: Multi-lane concurrency support for sub-agents via sessions_spawn</td><td>0.5000</td><td>queueing, sessions, coding_agents</td><td>queueing, sessions</td><td>none</td><td>coding_agents</td></tr><tr><td><a href='https://github.com/openclaw/openclaw/issues/82507'>#82507</a></td><td>[Feature]: ACPX Codex sandbox should inherit user-installed plugins (e.g. Superpowers)</td><td>0.5000</td><td>acpx, codex, skills_plugins</td><td>acpx, skills_plugins</td><td>none</td><td>codex</td></tr><tr><td><a href='https://github.com/openclaw/openclaw/issues/40332'>#40332</a></td><td>[Feature]: Per-binding and per-agent permissionMode for ACP sessions</td><td>0.5000</td><td>acp, approvals, acpx</td><td>approvals, acpx</td><td>none</td><td>acp</td></tr><tr><td><a href='https://github.com/openclaw/openclaw/pull/84670'>#84670</a></td><td>[codex] fix webchat full-message reader for truncated history</td><td>0.5000</td><td>gateway, api_surface, ui_tui</td><td>gateway, ui_tui</td><td>none</td><td>api_surface</td></tr><tr><td><a href='https://github.com/openclaw/openclaw/issues/84583'>#84583</a></td><td>cron announce delivery triggers EmbeddedAttemptSessionTakeoverError when user is actively chatting</td><td>0.5000</td><td>cron_automation, sessions, reliability</td><td>cron_automation, sessions</td><td>none</td><td>reliability</td></tr><tr><td><a href='https://github.com/openclaw/openclaw/pull/68725'>#68725</a></td><td>feat(amazon-bedrock-mantle): add known context windows for open-weight Mantle models</td><td>0.5000</td><td>open_weight_models, local_model_providers</td><td>open_weight_models</td><td>none</td><td>local_model_providers</td></tr><tr><td><a href='https://github.com/openclaw/openclaw/pull/56442'>#56442</a></td><td>feat: Add opt-in ACP parent completion notify for sessions_spawn</td><td>0.5000</td><td>acp, sessions, agent_runtime</td><td>acp, sessions</td><td>none</td><td>agent_runtime</td></tr><tr><td><a href='https://github.com/openclaw/openclaw/issues/60979'>#60979</a></td><td>feature: sessions_spawn ACP delivery to channel (stream output to Zulip/Discord topic)</td><td>0.5000</td><td>acp, chat_integrations, sessions</td><td>acp, sessions</td><td>none</td><td>chat_integrations</td></tr><tr><td><a href='https://github.com/openclaw/openclaw/pull/52747'>#52747</a></td><td>fix(acp): time out stuck session lane tasks</td><td>0.5000</td><td>acp, sessions, reliability</td><td>acp, sessions</td><td>none</td><td>reliability</td></tr><tr><td><a href='https://github.com/openclaw/openclaw/pull/84763'>#84763</a></td><td>fix(acpx): scrub provider credential env from ACP harness spawns</td><td>0.5000</td><td>acpx, acp, security</td><td>acp, acpx</td><td>none</td><td>security</td></tr><tr><td><a href='https://github.com/openclaw/openclaw/pull/69256'>#69256</a></td><td>fix(cron): prevent premature session cleanup when subagents are running</td><td>0.5000</td><td>cron_automation, sessions, reliability</td><td>cron_automation, sessions</td><td>none</td><td>reliability</td></tr><tr><td><a href='https://github.com/openclaw/openclaw/pull/65242'>#65242</a></td><td>fix: CompletionDeliveryGate to prevent duplicate ACP completion delivery</td><td>0.5000</td><td>acp, coding_agents, reliability</td><td>acp, reliability</td><td>none</td><td>coding_agents</td></tr><tr><td><a href='https://github.com/openclaw/openclaw/pull/77827'>#77827</a></td><td>fix: LM Studio thinking blocks invisible with Responses API</td><td>0.5000</td><td>model_serving, local_models</td><td>local_models</td><td>none</td><td>model_serving</td></tr><tr><td><a href='https://github.com/openclaw/openclaw/pull/42027'>#42027</a></td><td>fix: resolve exec PATH fallback, layered browser diagnostics, and cron force-run deadlock</td><td>0.5000</td><td>exec_tools, browser_automation, cron_automation</td><td>exec_tools, browser_automation</td><td>none</td><td>cron_automation</td></tr><tr><td><a href='https://github.com/openclaw/openclaw/pull/84752'>#84752</a></td><td>fix: self-heal lane wedges + restore openai-codex OAuth on embedded path</td><td>0.5000</td><td>reliability, auth_identity, sessions</td><td>reliability, auth_identity</td><td>none</td><td>sessions</td></tr><tr><td><a href='https://github.com/openclaw/openclaw/pull/63826'>#63826</a></td><td>security: fix HIGH/CRITICAL vulns in skill scanner, SSRF, hook priority, and token verification</td><td>0.5000</td><td>security, hooks, skills_plugins</td><td>security, skills_plugins</td><td>none</td><td>hooks</td></tr><tr><td><a href='https://github.com/openclaw/openclaw/pull/62428'>#62428</a></td><td>test(exec): land exec v2 contract follow-through</td><td>0.5000</td><td>exec_tools, sandboxing, approvals</td><td>exec_tools, sandboxing</td><td>none</td><td>approvals</td></tr></tbody></table></div> | |
| </section> | |
| </main> | |
| </body> | |
| </html> | |