feat: add panic recovery with crash log & finish -> agent_exit#21
Merged
Conversation
…tter truncation Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
… cognitive reflection scope Co-Authored-By: Claude Code <noreply@anthropic.com>
…meter Co-Authored-By: Claude Code <noreply@anthropic.com>
Co-Authored-By: Claude Code <noreply@anthropic.com>
Reviewer's GuideIntroduces a centralized panic recovery with crash logging, renames and re-wires the core flow-control tool from Sequence diagram for panic recovery crash loggingsequenceDiagram
actor User
participant main
participant util_RecoverPanic as util.RecoverPanic
participant util_buildCrashReport as buildCrashReport
participant OS_FileSystem as OS_FileSystem
User->>main: start_application()
activate main
main->>main: defer util.RecoverPanic()
main->>main: execute_application_logic()
main-->>main: panic(recovered_value)
deactivate main
activate util_RecoverPanic
util_RecoverPanic->>util_RecoverPanic: r = recover()
alt r is nil
util_RecoverPanic-->>User: return (no crash)
else r is not nil
util_RecoverPanic->>util_buildCrashReport: buildCrashReport(r)
activate util_buildCrashReport
util_buildCrashReport-->>util_RecoverPanic: crash_report_string
deactivate util_buildCrashReport
util_RecoverPanic->>OS_FileSystem: crashLogDir() (resolve ~/.codeactor/logs/crash)
alt dir resolution fails
util_RecoverPanic->>User: write report to stderr
util_RecoverPanic-->>main: re_panic(r)
else dir ok
OS_FileSystem-->>util_RecoverPanic: crash_dir_path
util_RecoverPanic->>OS_FileSystem: MkdirAll(crash_dir_path)
alt MkdirAll fails
util_RecoverPanic->>User: write report to stderr
util_RecoverPanic-->>main: re_panic(r)
else MkdirAll ok
util_RecoverPanic->>OS_FileSystem: OpenFile(crash-YYYY-MM-DD.log, append)
alt OpenFile fails
util_RecoverPanic->>User: write report to stderr
util_RecoverPanic-->>main: re_panic(r)
else OpenFile ok
OS_FileSystem-->>util_RecoverPanic: file_handle
util_RecoverPanic->>OS_FileSystem: WriteString(report)
OS_FileSystem-->>util_RecoverPanic: write_result
util_RecoverPanic->>User: also print report to stderr
util_RecoverPanic-->>main: re_panic(r)
end
end
end
end
Updated class diagram for flow control, thinking tool, and agentsclassDiagram
class FlowControlTool {
+string WorkingDir
+ExecuteAgentExit(ctx context.Context, params map_string_interface) (interface_type, error)
}
class ThinkingTool {
+Name() string
+Call(ctx context.Context, input string) (string, error)
-buildCrashReport(recovered interface_type) string
}
class GlobalCtx {
+FlowOps *FlowOps
+ThinkingTool *ThinkingTool
}
class FlowOps {
+ExecuteAgentExit(ctx context.Context, params map_string_interface) (interface_type, error)
+ExecuteAskUserForHelp(ctx context.Context, params map_string_interface) (interface_type, error)
}
class ConductorAgent {
+GlobalCtx *GlobalCtx
+toolDefMap map[string]ToolDefinition
+getToolFunc(name string) tools.ToolFunc
+registerCustomAgent(ca *CustomAgent)
+Run(ctx context.Context, input string, mem *memory.Conversation) (string, error)
}
class CodingAgent {
+GlobalCtx *GlobalCtx
+NewCodingAgent(globalCtx *GlobalCtx, llm llms.LLM, maxSteps int) *CodingAgent
}
class ExecutorConfig {
+bool StopOnFinish
}
class RunAgentLoopFn {
+RunAgentLoop(ctx context.Context, cfg ExecutorConfig) (string, error)
}
%% Relationships
GlobalCtx --> FlowOps : has
GlobalCtx --> ThinkingTool : has
FlowOps --> FlowControlTool : uses
ConductorAgent --> GlobalCtx : has
CodingAgent --> GlobalCtx : has
ConductorAgent --> FlowOps : uses ExecuteAgentExit
CodingAgent --> FlowOps : uses ExecuteAgentExit
ConductorAgent --> ThinkingTool : uses Call
CodingAgent --> ThinkingTool : uses Call
ExecutorConfig --> RunAgentLoopFn : config_for
RunAgentLoopFn --> FlowOps : calls ExecuteAgentExit
class ToolDefinition {
+string Description
+map_string_interface Parameters
}
ConductorAgent --> ToolDefinition : stores_in_toolDefMap
File-Level Changes
Tips and commandsInteracting with Sourcery
Customizing Your ExperienceAccess your dashboard to:
Getting Help
|
![sourcery-ai[bot]](https://avatars.githubusercontent.com/in/48477?s=60&v=4){kind=small}
There was a problem hiding this comment.
Hey - I've found 1 issue, and left some high level feedback:
- In
FlowControlTool.ExecuteAgentExit, the error wrap context string is still"executeFinish", which is confusing now that the function and tool have been renamed; consider updating it to something like"executeAgentExit"for accurate diagnostics. - In
ThinkingTool.Call, the JSON field was renamed fromerror_messagetoproblem_descriptionbut the formatted output still labels it asError:; consider updating the label to match the new semantics to avoid confusion in logs.
Prompt for AI Agents
Please address the comments from this code review:
## Overall Comments
- In `FlowControlTool.ExecuteAgentExit`, the error wrap context string is still `"executeFinish"`, which is confusing now that the function and tool have been renamed; consider updating it to something like `"executeAgentExit"` for accurate diagnostics.
- In `ThinkingTool.Call`, the JSON field was renamed from `error_message` to `problem_description` but the formatted output still labels it as `Error:`; consider updating the label to match the new semantics to avoid confusion in logs.
## Individual Comments
### Comment 1
<location path="internal/agents/conductor.go" line_range="686" />
<code_context>
},
},
})
- if tc.FunctionCall.Name == "finish" {
+ if tc.FunctionCall.Name == "agent_exit" {
return "Task completed successfully", nil
}
</code_context>
<issue_to_address>
**issue (bug_risk):** The fixed "Task completed successfully" message no longer matches the broader semantics of agent_exit.
agent_exit can represent success, failure, clarification, or forced termination, but the returned string always says “Task completed successfully”. This risks misleading callers and any UI using this value. Consider deriving a status from the tool output (e.g., including the reason) or using a neutral message that doesn’t imply success.
</issue_to_address>Help me be more useful! Please click 👍 or 👎 on each comment and I'll use the feedback to improve your reviews.
| }, | ||
| }, | ||
| }) | ||
| if tc.FunctionCall.Name == "finish" { |
There was a problem hiding this comment.
issue (bug_risk): The fixed "Task completed successfully" message no longer matches the broader semantics of agent_exit.
agent_exit can represent success, failure, clarification, or forced termination, but the returned string always says “Task completed successfully”. This risks misleading callers and any UI using this value. Consider deriving a status from the tool output (e.g., including the reason) or using a neutral message that doesn’t imply success.
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 by Sourcery
Introduce robust crash logging and align agent flow-control tooling and TUI history presentation with the new exit semantics and more compact task titles.
New Features:
Enhancements:
finishtoagent_exitacross agents, executors, prompts, and tests to better represent agent termination semantics.problem_descriptionfield instead oferror_messagefor clearer logging semantics.Tests:
agent_exittool name and expectations instead offinish.