@devries-cameron20 / t3code Mirror · read-only / commits / afb8619e1d
Fork 0

The best way to control Claude Code, Codex and more from any computer or phone

Mirrored from https://github.com/Camo67/t3code.git

Synced Nov 22, 12:00 UTC

fix(pi): clear Pi's native queue before Stop aborts (#27514)

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com> Co-authored-by: GPT 6.1 Sol <codex@openai.com>

Wout Stiens committed Oct 30, 2026 at 17:09 UTC afb8619e1ddcda94388e3cba89f2c1384e85d711
4 files changed +141 -23
apps/server/src/orchestration-v2/testkit/fixtures/thread_rollback_after_stop/pi_transcript.ndjson
+6 -4
@@ -133,7 +133,9 @@
133 {"type":"emit_inbound","label":"tool_execution_start","frame":{"type":"tool_execution_start","toolCallId":"call_8ac3cb4177994abdb9aa1c57","toolName":"bash","args":{"command":"node -e \"console.log('interrupt fixture tool started'); setTimeout(() => {}, 30000)\"","timeout":35}}}
134 {"type":"emit_inbound","label":"tool_execution_update","frame":{"type":"tool_execution_update","toolCallId":"call_8ac3cb4177994abdb9aa1c57","toolName":"bash","args":{"command":"node -e \"console.log('interrupt fixture tool started'); setTimeout(() => {}, 30000)\"","timeout":35},"partialResult":{"content":[]}}}
135 {"type":"emit_inbound","label":"tool_execution_update","frame":{"type":"tool_execution_update","toolCallId":"call_8ac3cb4177994abdb9aa1c57","toolName":"bash","args":{"command":"node -e \"console.log('interrupt fixture tool started'); setTimeout(() => {}, 30000)\"","timeout":35},"partialResult":{"content":[{"type":"text","text":"interrupt fixture tool started\n"}],"details":{}}}}
136 -{"type":"expect_outbound","label":"abort","frame":{"type":"abort","id":"t3-9"}}
136 +{"type":"expect_outbound","label":"clear_queue","frame":{"type":"clear_queue","id":"t3-9"}}
137 +{"type":"emit_inbound","label":"response:clear_queue","frame":{"id":"t3-9","type":"response","command":"clear_queue","success":true,"data":{"steering":[],"followUp":[]}}}
138 +{"type":"expect_outbound","label":"abort","frame":{"type":"abort","id":"t3-10"}}
139 {"type":"emit_inbound","label":"tool_execution_end","frame":{"type":"tool_execution_end","toolCallId":"call_8ac3cb4177994abdb9aa1c57","toolName":"bash","result":{"content":[{"type":"text","text":"interrupt fixture tool started\n\n\nCommand aborted"}],"details":{}},"isError":true}}
140 {"type":"emit_inbound","label":"message_start","frame":{"type":"message_start","message":{"role":"toolResult","toolCallId":"call_8ac3cb4177994abdb9aa1c57","toolName":"bash","content":[{"type":"text","text":"interrupt fixture tool started\n\n\nCommand aborted"}],"details":{},"isError":true,"timestamp":1767225606463}}}
141 {"type":"emit_inbound","label":"message_end","frame":{"type":"message_end","message":{"role":"toolResult","toolCallId":"call_8ac3cb4177994abdb9aa1c57","toolName":"bash","content":[{"type":"text","text":"interrupt fixture tool started\n\n\nCommand aborted"}],"details":{},"isError":true,"timestamp":1767225606463}}}
@@ -144,9 +146,9 @@
146 {"type":"emit_inbound","label":"turn_end","frame":{"type":"turn_end","message":{"role":"assistant","content":[],"api":"openai-completions","provider":"openrouter","model":"deepseek/deepseek-v4-flash","usage":{"input":0,"output":0,"cacheRead":0,"cacheWrite":0,"totalTokens":0,"cost":{"input":0,"output":0,"cacheRead":0,"cacheWrite":0,"total":0}},"stopReason":"error","errorMessage":"This operation was aborted","timestamp":1767225606465,"thinkingLevel":"high"},"toolResults":[]}}
147 {"type":"emit_inbound","label":"agent_end","frame":{"type":"agent_end","messages":[{"role":"user","content":[{"type":"text","text":"Run this exact local command: `node -e \"console.log('interrupt fixture tool started'); setTimeout(() => {}, 30000)\"`. Do not answer until it completes, then respond exactly: interrupt fixture should not finish naturally."}],"timestamp":1767225601969},{"role":"assistant","content":[{"type":"thinking","thinking":"The user wants me to run a command that will wait for 30 seconds, and then after it completes, respond with a specific message. But they also say \"Do not answer until it completes\" - but this command will run for 30 seconds and then naturally finish. However, the response they want is \"interrupt fixture should not finish naturally\" which contradicts the fact that it did finish naturally.\n\nWait, let me re-read. They want me to run the command and not answer until it completes. The command is a node script that logs \"interrupt fixture tool started\" and then waits 30 seconds. If I wait for it to complete, it will finish naturally after 30 seconds. But the response they want is \"interrupt fixture should not finish naturally\" - implying that the fixture should have been interrupted before finishing naturally.\n\nThis is a test scenario. Let me just run it and see what happens. Actually, maybe they want me to just run it and respond accordingly. Let me execute the command.\n\nActually, I think this is part of a test where the system might interrupt/kill the process. Let me just run the command with a sufficient timeout.","thinkingSignature":"<signature>"},{"type":"toolCall","id":"call_8ac3cb4177994abdb9aa1c57","name":"bash","arguments":{"command":"node -e \"console.log('interrupt fixture tool started'); setTimeout(() => {}, 30000)\"","timeout":35}}],"api":"openai-completions","provider":"openrouter","model":"deepseek/deepseek-v4-flash","usage":{"input":77,"output":317,"cacheRead":1792,"cacheWrite":0,"reasoning":284,"totalTokens":2186,"cost":{"input":0.00000322322,"output":0.000026539239999999998,"cacheRead":0.000015002623999999998,"cacheWrite":0,"total":0.000044765084}},"stopReason":"toolUse","timestamp":1767225601972,"responseId":"gen-1790898593-E5hfYfRp6aKBI57pUKss","rawStopReason":"tool_calls","thinkingLevel":"high"},{"role":"toolResult","toolCallId":"call_8ac3cb4177994abdb9aa1c57","toolName":"bash","content":[{"type":"text","text":"interrupt fixture tool started\n\n\nCommand aborted"}],"details":{},"isError":true,"timestamp":1767225606463},{"role":"assistant","content":[],"api":"openai-completions","provider":"openrouter","model":"deepseek/deepseek-v4-flash","usage":{"input":0,"output":0,"cacheRead":0,"cacheWrite":0,"totalTokens":0,"cost":{"input":0,"output":0,"cacheRead":0,"cacheWrite":0,"total":0}},"stopReason":"error","errorMessage":"This operation was aborted","timestamp":1767225606465,"thinkingLevel":"high"}],"willRetry":false}}
148 {"type":"emit_inbound","label":"agent_settled","frame":{"type":"agent_settled"}}
147 -{"type":"emit_inbound","label":"response:abort","frame":{"id":"t3-9","type":"response","command":"abort","success":true}}
148 -{"type":"expect_outbound","label":"get_entries","frame":{"type":"get_entries","since":"b184b5a8","id":"t3-10"}}
149 -{"type":"emit_inbound","label":"response:get_entries","frame":{"id":"t3-10","type":"response","command":"get_entries","success":true,"data":{"entries":[{"type":"message","id":"ee73daae","parentId":"b184b5a8","timestamp":"2026-01-01T00:00:01.971Z","message":{"role":"user","content":[{"type":"text","text":"Run this exact local command: `node -e \"console.log('interrupt fixture tool started'); setTimeout(() => {}, 30000)\"`. Do not answer until it completes, then respond exactly: interrupt fixture should not finish naturally."}],"timestamp":1767225601969}},{"type":"message","id":"c3915df7","parentId":"ee73daae","timestamp":"2026-01-01T00:00:06.406Z","message":{"role":"assistant","content":[{"type":"thinking","thinking":"The user wants me to run a command that will wait for 30 seconds, and then after it completes, respond with a specific message. But they also say \"Do not answer until it completes\" - but this command will run for 30 seconds and then naturally finish. However, the response they want is \"interrupt fixture should not finish naturally\" which contradicts the fact that it did finish naturally.\n\nWait, let me re-read. They want me to run the command and not answer until it completes. The command is a node script that logs \"interrupt fixture tool started\" and then waits 30 seconds. If I wait for it to complete, it will finish naturally after 30 seconds. But the response they want is \"interrupt fixture should not finish naturally\" - implying that the fixture should have been interrupted before finishing naturally.\n\nThis is a test scenario. Let me just run it and see what happens. Actually, maybe they want me to just run it and respond accordingly. Let me execute the command.\n\nActually, I think this is part of a test where the system might interrupt/kill the process. Let me just run the command with a sufficient timeout.","thinkingSignature":"<signature>"},{"type":"toolCall","id":"call_8ac3cb4177994abdb9aa1c57","name":"bash","arguments":{"command":"node -e \"console.log('interrupt fixture tool started'); setTimeout(() => {}, 30000)\"","timeout":35}}],"api":"openai-completions","provider":"openrouter","model":"deepseek/deepseek-v4-flash","usage":{"input":77,"output":317,"cacheRead":1792,"cacheWrite":0,"reasoning":284,"totalTokens":2186,"cost":{"input":0.00000322322,"output":0.000026539239999999998,"cacheRead":0.000015002623999999998,"cacheWrite":0,"total":0.000044765084}},"stopReason":"toolUse","timestamp":1767225601972,"responseId":"gen-1790898593-E5hfYfRp6aKBI57pUKss","rawStopReason":"tool_calls","thinkingLevel":"high"}},{"type":"message","id":"641f3d3f","parentId":"c3915df7","timestamp":"2026-01-01T00:00:06.463Z","message":{"role":"toolResult","toolCallId":"call_8ac3cb4177994abdb9aa1c57","toolName":"bash","content":[{"type":"text","text":"interrupt fixture tool started\n\n\nCommand aborted"}],"details":{},"isError":true,"timestamp":1767225606463}},{"type":"message","id":"e3ab1495","parentId":"641f3d3f","timestamp":"2026-01-01T00:00:06.465Z","message":{"role":"assistant","content":[],"api":"openai-completions","provider":"openrouter","model":"deepseek/deepseek-v4-flash","usage":{"input":0,"output":0,"cacheRead":0,"cacheWrite":0,"totalTokens":0,"cost":{"input":0,"output":0,"cacheRead":0,"cacheWrite":0,"total":0}},"stopReason":"error","errorMessage":"This operation was aborted","timestamp":1767225606465,"thinkingLevel":"high"}}],"leafId":"e3ab1495"}}}
149 +{"type":"emit_inbound","label":"response:abort","frame":{"id":"t3-10","type":"response","command":"abort","success":true}}
150 +{"type":"expect_outbound","label":"get_entries","frame":{"type":"get_entries","since":"b184b5a8","id":"t3-11"}}
151 +{"type":"emit_inbound","label":"response:get_entries","frame":{"id":"t3-11","type":"response","command":"get_entries","success":true,"data":{"entries":[{"type":"message","id":"ee73daae","parentId":"b184b5a8","timestamp":"2026-01-01T00:00:01.971Z","message":{"role":"user","content":[{"type":"text","text":"Run this exact local command: `node -e \"console.log('interrupt fixture tool started'); setTimeout(() => {}, 30000)\"`. Do not answer until it completes, then respond exactly: interrupt fixture should not finish naturally."}],"timestamp":1767225601969}},{"type":"message","id":"c3915df7","parentId":"ee73daae","timestamp":"2026-01-01T00:00:06.406Z","message":{"role":"assistant","content":[{"type":"thinking","thinking":"The user wants me to run a command that will wait for 30 seconds, and then after it completes, respond with a specific message. But they also say \"Do not answer until it completes\" - but this command will run for 30 seconds and then naturally finish. However, the response they want is \"interrupt fixture should not finish naturally\" which contradicts the fact that it did finish naturally.\n\nWait, let me re-read. They want me to run the command and not answer until it completes. The command is a node script that logs \"interrupt fixture tool started\" and then waits 30 seconds. If I wait for it to complete, it will finish naturally after 30 seconds. But the response they want is \"interrupt fixture should not finish naturally\" - implying that the fixture should have been interrupted before finishing naturally.\n\nThis is a test scenario. Let me just run it and see what happens. Actually, maybe they want me to just run it and respond accordingly. Let me execute the command.\n\nActually, I think this is part of a test where the system might interrupt/kill the process. Let me just run the command with a sufficient timeout.","thinkingSignature":"<signature>"},{"type":"toolCall","id":"call_8ac3cb4177994abdb9aa1c57","name":"bash","arguments":{"command":"node -e \"console.log('interrupt fixture tool started'); setTimeout(() => {}, 30000)\"","timeout":35}}],"api":"openai-completions","provider":"openrouter","model":"deepseek/deepseek-v4-flash","usage":{"input":77,"output":317,"cacheRead":1792,"cacheWrite":0,"reasoning":284,"totalTokens":2186,"cost":{"input":0.00000322322,"output":0.000026539239999999998,"cacheRead":0.000015002623999999998,"cacheWrite":0,"total":0.000044765084}},"stopReason":"toolUse","timestamp":1767225601972,"responseId":"gen-1790898593-E5hfYfRp6aKBI57pUKss","rawStopReason":"tool_calls","thinkingLevel":"high"}},{"type":"message","id":"641f3d3f","parentId":"c3915df7","timestamp":"2026-01-01T00:00:06.463Z","message":{"role":"toolResult","toolCallId":"call_8ac3cb4177994abdb9aa1c57","toolName":"bash","content":[{"type":"text","text":"interrupt fixture tool started\n\n\nCommand aborted"}],"details":{},"isError":true,"timestamp":1767225606463}},{"type":"message","id":"e3ab1495","parentId":"641f3d3f","timestamp":"2026-01-01T00:00:06.465Z","message":{"role":"assistant","content":[],"api":"openai-completions","provider":"openrouter","model":"deepseek/deepseek-v4-flash","usage":{"input":0,"output":0,"cacheRead":0,"cacheWrite":0,"totalTokens":0,"cost":{"input":0,"output":0,"cacheRead":0,"cacheWrite":0,"total":0}},"stopReason":"error","errorMessage":"This operation was aborted","timestamp":1767225606465,"thinkingLevel":"high"}}],"leafId":"e3ab1495"}}}
152 {"type":"expect_outbound","label":"process_start@p2","frame":{"type":"process_start","args":["--mode","rpc","--provider","openrouter","--model","deepseek/deepseek-v4-flash","--no-extensions","--no-skills","--no-prompt-templates","--no-context-files","--approve","--extension","<any>"]}}
153 {"type":"expect_outbound","label":"switch_session@p2","frame":{"type":"switch_session","sessionPath":"/pi-sessions/session-1.jsonl","id":"t3-0"}}
154 {"type":"emit_inbound","label":"extension_ui_request@p2","frame":{"type":"extension_ui_request","id":"00000000-0000-4000-8000-000000000003","method":"notify","message":"t3-code MCP unavailable: T3_MCP_URL or T3_MCP_BEARER_TOKEN is missing.","notifyType":"warning"}}
apps/server/src/orchestration-v2/testkit/fixtures/turn_interrupt_mid_tool/pi_transcript.ndjson
+6 -4
@@ -41,7 +41,9 @@
41 {"type":"emit_inbound","label":"message_end","frame":{"type":"message_end","message":{"role":"assistant","content":[{"type":"thinking","thinking":"The user wants me to run a specific command and wait for it to complete. Let me execute it.","thinkingSignature":"<signature>"},{"type":"toolCall","id":"call_9726ad99cc114832996228eb","name":"bash","arguments":{"command":"node -e \"console.log('interrupt fixture tool started'); setTimeout(() => {}, 30000)\"","timeout":35}}],"api":"openai-completions","provider":"openrouter","model":"deepseek/deepseek-v4-flash","usage":{"input":1830,"output":102,"cacheRead":0,"cacheWrite":0,"reasoning":23,"totalTokens":1932,"cost":{"input":0.0000766038,"output":0.00000853944,"cacheRead":0,"cacheWrite":0,"total":0.00008514324}},"stopReason":"toolUse","timestamp":1767225600401,"responseId":"gen-1790898592-Mzq78XbaTGiNE3YSafgy","rawStopReason":"tool_calls","thinkingLevel":"high"}}}
42 {"type":"emit_inbound","label":"tool_execution_start","frame":{"type":"tool_execution_start","toolCallId":"call_9726ad99cc114832996228eb","toolName":"bash","args":{"command":"node -e \"console.log('interrupt fixture tool started'); setTimeout(() => {}, 30000)\"","timeout":35}}}
43 {"type":"emit_inbound","label":"tool_execution_update","frame":{"type":"tool_execution_update","toolCallId":"call_9726ad99cc114832996228eb","toolName":"bash","args":{"command":"node -e \"console.log('interrupt fixture tool started'); setTimeout(() => {}, 30000)\"","timeout":35},"partialResult":{"content":[]}}}
44 -{"type":"expect_outbound","label":"abort","frame":{"type":"abort","id":"t3-6"}}
44 +{"type":"expect_outbound","label":"clear_queue","frame":{"type":"clear_queue","id":"t3-6"}}
45 +{"type":"emit_inbound","label":"response:clear_queue","frame":{"id":"t3-6","type":"response","command":"clear_queue","success":true,"data":{"steering":[],"followUp":[]}}}
46 +{"type":"expect_outbound","label":"abort","frame":{"type":"abort","id":"t3-7"}}
47 {"type":"emit_inbound","label":"tool_execution_end","frame":{"type":"tool_execution_end","toolCallId":"call_9726ad99cc114832996228eb","toolName":"bash","result":{"content":[{"type":"text","text":"Command aborted"}],"details":{}},"isError":true}}
48 {"type":"emit_inbound","label":"message_start","frame":{"type":"message_start","message":{"role":"toolResult","toolCallId":"call_9726ad99cc114832996228eb","toolName":"bash","content":[{"type":"text","text":"Command aborted"}],"details":{},"isError":true,"timestamp":1767225602282}}}
49 {"type":"emit_inbound","label":"message_end","frame":{"type":"message_end","message":{"role":"toolResult","toolCallId":"call_9726ad99cc114832996228eb","toolName":"bash","content":[{"type":"text","text":"Command aborted"}],"details":{},"isError":true,"timestamp":1767225602282}}}
@@ -52,6 +54,6 @@
54 {"type":"emit_inbound","label":"turn_end","frame":{"type":"turn_end","message":{"role":"assistant","content":[],"api":"openai-completions","provider":"openrouter","model":"deepseek/deepseek-v4-flash","usage":{"input":0,"output":0,"cacheRead":0,"cacheWrite":0,"totalTokens":0,"cost":{"input":0,"output":0,"cacheRead":0,"cacheWrite":0,"total":0}},"stopReason":"error","errorMessage":"This operation was aborted","timestamp":1767225602287,"thinkingLevel":"high"},"toolResults":[]}}
55 {"type":"emit_inbound","label":"agent_end","frame":{"type":"agent_end","messages":[{"role":"system","content":"","sections":{"preamble":"<system-prompt-section>","tools":"<system-prompt-section>","rules":"<system-prompt-section>","docs":"<system-prompt-section>","cwd":"<system-prompt-section>"},"timestamp":1790898591817,"toolsAdded":[{"name":"read","description":"Read the contents of a file. Supports text files and images (jpg, png, gif, webp, bmp). Images are sent as attachments. For text files, output is truncated to 2000 lines or 50KB (whichever is hit first). Use offset/limit for large files. When you need the full file, continue with offset until complete.","parameters":{"type":"object","required":["path"],"properties":{"path":{"type":"string","description":"Path to the file to read (relative or absolute)"},"offset":{"type":"number","description":"Line number to start reading from (1-indexed)"},"limit":{"type":"number","description":"Maximum number of lines to read"}}},"constrainedSampling":{"type":"json_schema","strict":"prefer"}},{"name":"bash","description":"Execute a bash command in the current working directory. Returns stdout and stderr. Output is truncated to last 2000 lines or 50KB (whichever is hit first). If truncated, full output is saved to a temp file. Optionally provide a timeout in seconds.","parameters":{"type":"object","required":["command"],"properties":{"command":{"type":"string","description":"Shell command to execute"},"timeout":{"type":"number","description":"Timeout in seconds (optional, no default timeout)"}}},"constrainedSampling":{"type":"json_schema","strict":"prefer"}},{"name":"edit","description":"Edit a single file using exact text replacement. Every edits[].oldText must match a unique, non-overlapping region of the original file. If two changes affect the same block or nearby lines, merge them into one edit instead of emitting overlapping edits. Do not include large unchanged regions just to connect distant changes.","parameters":{"type":"object","required":["path","edits"],"properties":{"path":{"type":"string","description":"Path to the file to edit (relative or absolute)"},"edits":{"type":"array","items":{"type":"object","required":["oldText","newText"],"properties":{"oldText":{"type":"string","description":"Exact text for one targeted replacement. It must be unique in the original file and must not overlap with any other edits[].oldText in the same call."},"newText":{"type":"string","description":"Replacement text for this targeted edit."}}},"description":"One or more targeted replacements. Each edit is matched against the original file, not incrementally. Do not include overlapping or nested edits. If two changes touch the same block or nearby lines, merge them into one edit instead."}}},"constrainedSampling":{"type":"json_schema","strict":"prefer"}},{"name":"write","description":"Write content to a file. Creates the file if it doesn't exist, overwrites if it does. Automatically creates parent directories.","parameters":{"type":"object","required":["path","content"],"properties":{"path":{"type":"string","description":"Path to the file to write (relative or absolute)"},"content":{"type":"string","description":"Content to write to the file"}}},"constrainedSampling":{"type":"json_schema","strict":"prefer"}}]},{"role":"user","content":[{"type":"text","text":"Run this exact local command: `node -e \"console.log('interrupt fixture tool started'); setTimeout(() => {}, 30000)\"`. Do not answer until it completes, then respond exactly: interrupt fixture should not finish naturally."}],"timestamp":1767225600375},{"role":"assistant","content":[{"type":"thinking","thinking":"The user wants me to run a specific command and wait for it to complete. Let me execute it.","thinkingSignature":"<signature>"},{"type":"toolCall","id":"call_9726ad99cc114832996228eb","name":"bash","arguments":{"command":"node -e \"console.log('interrupt fixture tool started'); setTimeout(() => {}, 30000)\"","timeout":35}}],"api":"openai-completions","provider":"openrouter","model":"deepseek/deepseek-v4-flash","usage":{"input":1830,"output":102,"cacheRead":0,"cacheWrite":0,"reasoning":23,"totalTokens":1932,"cost":{"input":0.0000766038,"output":0.00000853944,"cacheRead":0,"cacheWrite":0,"total":0.00008514324}},"stopReason":"toolUse","timestamp":1767225600401,"responseId":"gen-1790898592-Mzq78XbaTGiNE3YSafgy","rawStopReason":"tool_calls","thinkingLevel":"high"},{"role":"toolResult","toolCallId":"call_9726ad99cc114832996228eb","toolName":"bash","content":[{"type":"text","text":"Command aborted"}],"details":{},"isError":true,"timestamp":1767225602282},{"role":"assistant","content":[],"api":"openai-completions","provider":"openrouter","model":"deepseek/deepseek-v4-flash","usage":{"input":0,"output":0,"cacheRead":0,"cacheWrite":0,"totalTokens":0,"cost":{"input":0,"output":0,"cacheRead":0,"cacheWrite":0,"total":0}},"stopReason":"error","errorMessage":"This operation was aborted","timestamp":1767225602287,"thinkingLevel":"high"}],"willRetry":false}}
56 {"type":"emit_inbound","label":"agent_settled","frame":{"type":"agent_settled"}}
55 -{"type":"emit_inbound","label":"response:abort","frame":{"id":"t3-6","type":"response","command":"abort","success":true}}
56 -{"type":"expect_outbound","label":"get_entries","frame":{"type":"get_entries","since":"4405fe6a","id":"t3-7"}}
57 -{"type":"emit_inbound","label":"response:get_entries","frame":{"id":"t3-7","type":"response","command":"get_entries","success":true,"data":{"entries":[{"type":"model_change","id":"b78c2012","parentId":"4405fe6a","timestamp":"2026-01-01T00:00:00.366Z","provider":"openrouter","modelId":"deepseek/deepseek-v4-flash"},{"type":"session_info","id":"9dea9c23","parentId":"b78c2012","timestamp":"2026-01-01T00:00:00.370Z","name":"Replay fixture: turn_interrupt_mid_tool"},{"type":"message","id":"8dc98fa4","parentId":"9dea9c23","timestamp":"2026-01-01T00:00:00.379Z","message":{"role":"system","content":"","sections":{"preamble":"<system-prompt-section>","tools":"<system-prompt-section>","rules":"<system-prompt-section>","docs":"<system-prompt-section>","cwd":"<system-prompt-section>"},"timestamp":1790898591817,"toolsAdded":[{"name":"read","description":"Read the contents of a file. Supports text files and images (jpg, png, gif, webp, bmp). Images are sent as attachments. For text files, output is truncated to 2000 lines or 50KB (whichever is hit first). Use offset/limit for large files. When you need the full file, continue with offset until complete.","parameters":{"type":"object","required":["path"],"properties":{"path":{"type":"string","description":"Path to the file to read (relative or absolute)"},"offset":{"type":"number","description":"Line number to start reading from (1-indexed)"},"limit":{"type":"number","description":"Maximum number of lines to read"}}},"constrainedSampling":{"type":"json_schema","strict":"prefer"}},{"name":"bash","description":"Execute a bash command in the current working directory. Returns stdout and stderr. Output is truncated to last 2000 lines or 50KB (whichever is hit first). If truncated, full output is saved to a temp file. Optionally provide a timeout in seconds.","parameters":{"type":"object","required":["command"],"properties":{"command":{"type":"string","description":"Shell command to execute"},"timeout":{"type":"number","description":"Timeout in seconds (optional, no default timeout)"}}},"constrainedSampling":{"type":"json_schema","strict":"prefer"}},{"name":"edit","description":"Edit a single file using exact text replacement. Every edits[].oldText must match a unique, non-overlapping region of the original file. If two changes affect the same block or nearby lines, merge them into one edit instead of emitting overlapping edits. Do not include large unchanged regions just to connect distant changes.","parameters":{"type":"object","required":["path","edits"],"properties":{"path":{"type":"string","description":"Path to the file to edit (relative or absolute)"},"edits":{"type":"array","items":{"type":"object","required":["oldText","newText"],"properties":{"oldText":{"type":"string","description":"Exact text for one targeted replacement. It must be unique in the original file and must not overlap with any other edits[].oldText in the same call."},"newText":{"type":"string","description":"Replacement text for this targeted edit."}}},"description":"One or more targeted replacements. Each edit is matched against the original file, not incrementally. Do not include overlapping or nested edits. If two changes touch the same block or nearby lines, merge them into one edit instead."}}},"constrainedSampling":{"type":"json_schema","strict":"prefer"}},{"name":"write","description":"Write content to a file. Creates the file if it doesn't exist, overwrites if it does. Automatically creates parent directories.","parameters":{"type":"object","required":["path","content"],"properties":{"path":{"type":"string","description":"Path to the file to write (relative or absolute)"},"content":{"type":"string","description":"Content to write to the file"}}},"constrainedSampling":{"type":"json_schema","strict":"prefer"}}]}},{"type":"message","id":"b36f8c4a","parentId":"8dc98fa4","timestamp":"2026-01-01T00:00:00.379Z","message":{"role":"user","content":[{"type":"text","text":"Run this exact local command: `node -e \"console.log('interrupt fixture tool started'); setTimeout(() => {}, 30000)\"`. Do not answer until it completes, then respond exactly: interrupt fixture should not finish naturally."}],"timestamp":1767225600375}},{"type":"message","id":"8ea47a23","parentId":"b36f8c4a","timestamp":"2026-01-01T00:00:02.246Z","message":{"role":"assistant","content":[{"type":"thinking","thinking":"The user wants me to run a specific command and wait for it to complete. Let me execute it.","thinkingSignature":"<signature>"},{"type":"toolCall","id":"call_9726ad99cc114832996228eb","name":"bash","arguments":{"command":"node -e \"console.log('interrupt fixture tool started'); setTimeout(() => {}, 30000)\"","timeout":35}}],"api":"openai-completions","provider":"openrouter","model":"deepseek/deepseek-v4-flash","usage":{"input":1830,"output":102,"cacheRead":0,"cacheWrite":0,"reasoning":23,"totalTokens":1932,"cost":{"input":0.0000766038,"output":0.00000853944,"cacheRead":0,"cacheWrite":0,"total":0.00008514324}},"stopReason":"toolUse","timestamp":1767225600401,"responseId":"gen-1790898592-Mzq78XbaTGiNE3YSafgy","rawStopReason":"tool_calls","thinkingLevel":"high"}},{"type":"message","id":"aa5aec89","parentId":"8ea47a23","timestamp":"2026-01-01T00:00:02.282Z","message":{"role":"toolResult","toolCallId":"call_9726ad99cc114832996228eb","toolName":"bash","content":[{"type":"text","text":"Command aborted"}],"details":{},"isError":true,"timestamp":1767225602282}},{"type":"message","id":"206b2637","parentId":"aa5aec89","timestamp":"2026-01-01T00:00:02.287Z","message":{"role":"assistant","content":[],"api":"openai-completions","provider":"openrouter","model":"deepseek/deepseek-v4-flash","usage":{"input":0,"output":0,"cacheRead":0,"cacheWrite":0,"totalTokens":0,"cost":{"input":0,"output":0,"cacheRead":0,"cacheWrite":0,"total":0}},"stopReason":"error","errorMessage":"This operation was aborted","timestamp":1767225602287,"thinkingLevel":"high"}}],"leafId":"206b2637"}}}
57 +{"type":"emit_inbound","label":"response:abort","frame":{"id":"t3-7","type":"response","command":"abort","success":true}}
58 +{"type":"expect_outbound","label":"get_entries","frame":{"type":"get_entries","since":"4405fe6a","id":"t3-8"}}
59 +{"type":"emit_inbound","label":"response:get_entries","frame":{"id":"t3-8","type":"response","command":"get_entries","success":true,"data":{"entries":[{"type":"model_change","id":"b78c2012","parentId":"4405fe6a","timestamp":"2026-01-01T00:00:00.366Z","provider":"openrouter","modelId":"deepseek/deepseek-v4-flash"},{"type":"session_info","id":"9dea9c23","parentId":"b78c2012","timestamp":"2026-01-01T00:00:00.370Z","name":"Replay fixture: turn_interrupt_mid_tool"},{"type":"message","id":"8dc98fa4","parentId":"9dea9c23","timestamp":"2026-01-01T00:00:00.379Z","message":{"role":"system","content":"","sections":{"preamble":"<system-prompt-section>","tools":"<system-prompt-section>","rules":"<system-prompt-section>","docs":"<system-prompt-section>","cwd":"<system-prompt-section>"},"timestamp":1790898591817,"toolsAdded":[{"name":"read","description":"Read the contents of a file. Supports text files and images (jpg, png, gif, webp, bmp). Images are sent as attachments. For text files, output is truncated to 2000 lines or 50KB (whichever is hit first). Use offset/limit for large files. When you need the full file, continue with offset until complete.","parameters":{"type":"object","required":["path"],"properties":{"path":{"type":"string","description":"Path to the file to read (relative or absolute)"},"offset":{"type":"number","description":"Line number to start reading from (1-indexed)"},"limit":{"type":"number","description":"Maximum number of lines to read"}}},"constrainedSampling":{"type":"json_schema","strict":"prefer"}},{"name":"bash","description":"Execute a bash command in the current working directory. Returns stdout and stderr. Output is truncated to last 2000 lines or 50KB (whichever is hit first). If truncated, full output is saved to a temp file. Optionally provide a timeout in seconds.","parameters":{"type":"object","required":["command"],"properties":{"command":{"type":"string","description":"Shell command to execute"},"timeout":{"type":"number","description":"Timeout in seconds (optional, no default timeout)"}}},"constrainedSampling":{"type":"json_schema","strict":"prefer"}},{"name":"edit","description":"Edit a single file using exact text replacement. Every edits[].oldText must match a unique, non-overlapping region of the original file. If two changes affect the same block or nearby lines, merge them into one edit instead of emitting overlapping edits. Do not include large unchanged regions just to connect distant changes.","parameters":{"type":"object","required":["path","edits"],"properties":{"path":{"type":"string","description":"Path to the file to edit (relative or absolute)"},"edits":{"type":"array","items":{"type":"object","required":["oldText","newText"],"properties":{"oldText":{"type":"string","description":"Exact text for one targeted replacement. It must be unique in the original file and must not overlap with any other edits[].oldText in the same call."},"newText":{"type":"string","description":"Replacement text for this targeted edit."}}},"description":"One or more targeted replacements. Each edit is matched against the original file, not incrementally. Do not include overlapping or nested edits. If two changes touch the same block or nearby lines, merge them into one edit instead."}}},"constrainedSampling":{"type":"json_schema","strict":"prefer"}},{"name":"write","description":"Write content to a file. Creates the file if it doesn't exist, overwrites if it does. Automatically creates parent directories.","parameters":{"type":"object","required":["path","content"],"properties":{"path":{"type":"string","description":"Path to the file to write (relative or absolute)"},"content":{"type":"string","description":"Content to write to the file"}}},"constrainedSampling":{"type":"json_schema","strict":"prefer"}}]}},{"type":"message","id":"b36f8c4a","parentId":"8dc98fa4","timestamp":"2026-01-01T00:00:00.379Z","message":{"role":"user","content":[{"type":"text","text":"Run this exact local command: `node -e \"console.log('interrupt fixture tool started'); setTimeout(() => {}, 30000)\"`. Do not answer until it completes, then respond exactly: interrupt fixture should not finish naturally."}],"timestamp":1767225600375}},{"type":"message","id":"8ea47a23","parentId":"b36f8c4a","timestamp":"2026-01-01T00:00:02.246Z","message":{"role":"assistant","content":[{"type":"thinking","thinking":"The user wants me to run a specific command and wait for it to complete. Let me execute it.","thinkingSignature":"<signature>"},{"type":"toolCall","id":"call_9726ad99cc114832996228eb","name":"bash","arguments":{"command":"node -e \"console.log('interrupt fixture tool started'); setTimeout(() => {}, 30000)\"","timeout":35}}],"api":"openai-completions","provider":"openrouter","model":"deepseek/deepseek-v4-flash","usage":{"input":1830,"output":102,"cacheRead":0,"cacheWrite":0,"reasoning":23,"totalTokens":1932,"cost":{"input":0.0000766038,"output":0.00000853944,"cacheRead":0,"cacheWrite":0,"total":0.00008514324}},"stopReason":"toolUse","timestamp":1767225600401,"responseId":"gen-1790898592-Mzq78XbaTGiNE3YSafgy","rawStopReason":"tool_calls","thinkingLevel":"high"}},{"type":"message","id":"aa5aec89","parentId":"8ea47a23","timestamp":"2026-01-01T00:00:02.282Z","message":{"role":"toolResult","toolCallId":"call_9726ad99cc114832996228eb","toolName":"bash","content":[{"type":"text","text":"Command aborted"}],"details":{},"isError":true,"timestamp":1767225602282}},{"type":"message","id":"206b2637","parentId":"aa5aec89","timestamp":"2026-01-01T00:00:02.287Z","message":{"role":"assistant","content":[],"api":"openai-completions","provider":"openrouter","model":"deepseek/deepseek-v4-flash","usage":{"input":0,"output":0,"cacheRead":0,"cacheWrite":0,"totalTokens":0,"cost":{"input":0,"output":0,"cacheRead":0,"cacheWrite":0,"total":0}},"stopReason":"error","errorMessage":"This operation was aborted","timestamp":1767225602287,"thinkingLevel":"high"}}],"leafId":"206b2637"}}}
packages/provider-pi/src/server/adapter.test.ts
+69
@@ -169,6 +169,12 @@ interface FakePi {
169 readonly vetoNextNewSession: () => void;
170 readonly allRequests: () => ReadonlyArray<PiRpcRecord>;
171 readonly queueStats: (data: unknown) => void;
172 + /** Pi before 0.84.4 answers clear_queue as an unknown command. */
173 + readonly setClearQueue: (
174 + result:
175 + | { readonly steering: ReadonlyArray<string>; readonly followUp: ReadonlyArray<string> }
176 + | "unknown_command",
177 + ) => void;
178 readonly closeStdout: Effect.Effect<void>;
179 readonly lastSpawn: () => {
180 readonly args: ReadonlyArray<string>;
@@ -208,6 +214,7 @@ const makeFakePi: Effect.Effect<FakePi> = Effect.gen(function* () {
214 let sessionGeneration = 0;
215 let models: ReadonlyArray<unknown> = [];
216 let isStreaming: boolean | null = null;
217 + let clearQueue: Parameters<FakePi["setClearQueue"]>[0] = { steering: [], followUp: [] };
218 let stdinBuffer = "";
219
220 const emit = (record: PiRpcRecord, chunkBytes?: number) =>
@@ -273,6 +280,10 @@ const makeFakePi: Effect.Effect<FakePi> = Effect.gen(function* () {
280 return { ...base, data: statsQueue.shift() ?? {} };
281 case "fork":
282 return { ...base, data: { text: "Hello pi", cancelled: false } };
283 + case "clear_queue":
284 + return clearQueue === "unknown_command"
285 + ? { ...base, success: false, error: "Unknown command: clear_queue" }
286 + : { ...base, data: clearQueue };
287 default:
288 return base;
289 }
@@ -386,6 +397,9 @@ const makeFakePi: Effect.Effect<FakePi> = Effect.gen(function* () {
397 isStreaming = streaming;
398 },
399 queueStats: (data) => statsQueue.push(data),
400 + setClearQueue: (result) => {
401 + clearQueue = result;
402 + },
403 closeStdout: Queue.end(stdout),
404 lastSpawn: () => lastSpawn,
405 } satisfies FakePi;
@@ -3876,6 +3890,61 @@ describe("PiAdapterV2", () => {
3890 }).pipe(Effect.scoped, Effect.provide(layerTest)),
3891 );
3892
3893 + it.effect.each([
3894 + {
3895 + pi: "clears its native queue",
3896 + clearQueue: { steering: ["Use the other file", ""], followUp: ["Then summarize"] },
3897 + notices: [
3898 + "Stop cleared 3 queued messages Pi had not read:\n\nUse the other file\n\nThen summarize",
3899 + ],
3900 + },
3901 + { pi: "predates clear_queue", clearQueue: "unknown_command", notices: [] },
3902 + ] as const)("Stop aborts and retires Pi when it $pi", ({ clearQueue, notices }) =>
3903 + Effect.gen(function* () {
3904 + const fake = yield* makeFakePi;
3905 + fake.setClearQueue(clearQueue);
3906 + const { runtime, takeEvent } = yield* openRuntime(fake);
3907 + const providerThread = yield* runtime.ensureThread({
3908 + threadId: THREAD_ID,
3909 + modelSelection: modelSelection("default"),
3910 + runtimePolicy,
3911 + });
3912 + yield* startTurn(runtime, providerThread);
3913 + yield* fake.takeRequest("prompt");
3914 + yield* fake.emit({ type: "agent_start" });
3915 + const runningTurn = yield* takeEvent(
3916 + (event) =>
3917 + event.type === "provider_turn.updated" && event.providerTurn.status === "running",
3918 + );
3919 + const providerTurnId =
3920 + runningTurn.type === "provider_turn.updated" ? runningTurn.providerTurn.id : undefined;
3921 + assert.isDefined(providerTurnId);
3922 +
3923 + yield* runtime.interruptTurn({
3924 + providerThread,
3925 + providerTurnId: providerTurnId!,
3926 + requestRuntimeRestart: true,
3927 + });
3928 + yield* fake.takeRequest("abort");
3929 + assert.deepEqual(
3930 + fake
3931 + .allRequests()
3932 + .map((request) => request["type"])
3933 + .filter((type) => type === "clear_queue" || type === "abort"),
3934 + ["clear_queue", "abort"],
3935 + );
3936 + const emitted: Array<string> = [];
3937 + yield* takeEvent((event) => {
3938 + if (event.type === "turn_item.updated" && event.turnItem.type === "system_notice")
3939 + emitted.push(event.turnItem.message);
3940 + return (
3941 + event.type === "provider_session.updated" && event.providerSession.status === "stopped"
3942 + );
3943 + });
3944 + assert.deepEqual<ReadonlyArray<string>>(emitted, notices);
3945 + }).pipe(Effect.scoped, Effect.provide(layerTest)),
3946 + );
3947 +
3948 it.effect("stops active retry progress when the turn is interrupted", () =>
3949 Effect.gen(function* () {
3950 const fake = yield* makeFakePi;
packages/provider-pi/src/server/adapter.ts
+60 -15
@@ -1212,6 +1212,59 @@ export const makePiAdapterV2 = Effect.fn("makePiAdapterV2")(function* (
1212 });
1213 });
1214
1215 + const emitSystemNotice = Effect.fnUntraced(function* (
1216 + turn: ActivePiTurn,
1217 + nativeItemId: string,
1218 + message: string,
1219 + ) {
1220 + const emittedAt = yield* DateTime.now;
1221 + yield* emitItemNode(turn, nativeItemId, "system", "completed", emittedAt, emittedAt);
1222 + yield* emit({
1223 + type: "turn_item.updated",
1224 + driver: PI_PROVIDER,
1225 + turnItem: {
1226 + ...baseItemFields(turn, nativeItemId, emittedAt, emittedAt),
1227 + status: "completed",
1228 + completedAt: emittedAt,
1229 + title: message,
1230 + type: "system_notice",
1231 + message,
1232 + },
1233 + });
1234 + });
1235 +
1236 + // Abort leaves queued steers and follow-ups for Pi's next prompt, and Pi before 0.86 starts them
1237 + // as soon as abort lands, so Stop clears them first. Pi before 0.84.4 rejects clear_queue.
1238 + const clearNativeQueue = Effect.fnUntraced(function* (turn: ActivePiTurn) {
1239 + const cleared = yield* request({ type: "clear_queue" }, 2_000).pipe(
1240 + Effect.map((data) =>
1241 + ["steering", "followUp"].flatMap((key) => {
1242 + const texts = recordField(data, key);
1243 + return Array.isArray(texts)
1244 + ? texts.filter((text): text is string => typeof text === "string")
1245 + : [];
1246 + }),
1247 + ),
1248 + Effect.orElseSucceed((): ReadonlyArray<string> => []),
1249 + );
1250 + if (cleared.length === 0) return;
1251 + yield* sessionEventPermit.withPermits(1)(
1252 + threadState?.activeTurn === turn
1253 + ? emitSystemNotice(
1254 + turn,
1255 + `queue-cleared:${turn.providerTurn.id}`,
1256 + [
1257 + `Stop cleared ${cleared.length} queued ${cleared.length === 1 ? "message" : "messages"} Pi had not read:`,
1258 + // Image-only messages come back as empty text.
1259 + ...cleared.filter((text) => text.length > 0),
1260 + ]
1261 + .join("\n\n")
1262 + .slice(0, 2_000),
1263 + )
1264 + : Effect.void,
1265 + );
1266 + });
1267 +
1268 const handleExtensionUiRequest = Effect.fnUntraced(function* (event: PiRpcRecord) {
1269 const method = recordString(event, "method");
1270 const nativeRequestId = recordString(event, "id");
@@ -1221,21 +1274,11 @@ export const makePiAdapterV2 = Effect.fn("makePiAdapterV2")(function* (
1274 const turn = state?.activeTurn ?? null;
1275 const message = recordString(event, "message") ?? "";
1276 if (turn === null || message.length === 0) return;
1224 - const emittedAt = yield* DateTime.now;
1225 - const nativeItemId = `notify:${turn.providerTurn.id}:${turn.nextItemOrdinal}`;
1226 - yield* emitItemNode(turn, nativeItemId, "system", "completed", emittedAt, emittedAt);
1227 - yield* emit({
1228 - type: "turn_item.updated",
1229 - driver: PI_PROVIDER,
1230 - turnItem: {
1231 - ...baseItemFields(turn, nativeItemId, emittedAt, emittedAt),
1232 - status: "completed",
1233 - completedAt: emittedAt,
1234 - title: message,
1235 - type: "system_notice",
1236 - message,
1237 - },
1238 - });
1277 + yield* emitSystemNotice(
1278 + turn,
1279 + `notify:${turn.providerTurn.id}:${turn.nextItemOrdinal}`,
1280 + message,
1281 + );
1282 return;
1283 }
1284 if (
@@ -2641,6 +2684,7 @@ export const makePiAdapterV2 = Effect.fn("makePiAdapterV2")(function* (
2684 // Generic abort misses manual compaction; terminate so Stop covers /compact and recovery.
2685 stopRequested = true;
2686 if (interruptInput.requestRuntimeRestart === true && !turn.settleWhenIdle) {
2687 + yield* clearNativeQueue(turn);
2688 yield* request({ type: "abort" }, 2_000).pipe(Effect.ignore);
2689 }
2690 // Read rollback tree refs before termination breaks requests; the permit lets in-flight capture finish.
@@ -2654,6 +2698,7 @@ export const makePiAdapterV2 = Effect.fn("makePiAdapterV2")(function* (
2698 );
2699 return;
2700 }
2701 + yield* clearNativeQueue(turn);
2702 yield* request({ type: "abort" }).pipe(
2703 Effect.tapError(() => Effect.sync(() => (turn.interrupted = false))),
2704 );