-
-
Notifications
You must be signed in to change notification settings - Fork 0
feat(scripts): add error classification module with recovery hints #1331
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Merged
Merged
Changes from all commits
Commits
Show all changes
3 commits
Select commit
Hold shift + click to select a range
File filter
Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
There are no files selected for viewing
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
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,48 @@ | ||
| # Recovery hints for error classification (Issue #1330). | ||
| # | ||
| # Structure: top-level keys are sections. Each section is a list of | ||
| # {pattern, hint} entries. Patterns are case-insensitive regex. | ||
| # | ||
| # Naming convention: | ||
| # tool_<name> - hints scoped to a specific tool | ||
| # general - hints applied to all failures | ||
|
|
||
| tool_gh: | ||
| - pattern: "GraphQL: Could not resolve" | ||
| hint: "Issue or PR number may not exist. Verify with gh issue list or gh pr list." | ||
| - pattern: "HTTP 403" | ||
| hint: "Rate limited or insufficient permissions. Wait 60s or use gh api --cache." | ||
| - pattern: "HTTP 404" | ||
| hint: "Resource not found. Verify the owner/repo/number are correct." | ||
| - pattern: "HTTP 422" | ||
| hint: "Validation failed. Check required fields and value constraints." | ||
|
|
||
| tool_git: | ||
| - pattern: "not a git repository" | ||
| hint: "Current directory is not a git repo. Verify working directory." | ||
| - pattern: "CONFLICT" | ||
| hint: "Merge conflict detected. Resolve conflicts before continuing." | ||
| - pattern: "Permission denied" | ||
| hint: "File permission error. Check file ownership and access rights." | ||
|
|
||
| tool_python3: | ||
| - pattern: "ModuleNotFoundError" | ||
| hint: "Missing Python dependency. Run uv pip install or check pyproject.toml." | ||
| - pattern: "SyntaxError" | ||
| hint: "Python syntax error. Check the file for typos or version incompatibilities." | ||
|
|
||
| tool_npm: | ||
| - pattern: "ENOENT.*package\\.json" | ||
| hint: "No package.json found. Verify you are in the correct directory." | ||
| - pattern: "ERESOLVE" | ||
| hint: "Dependency resolution conflict. Try npm install --legacy-peer-deps." | ||
|
|
||
| general: | ||
| - pattern: "rate limit" | ||
| hint: "API rate limit hit. Wait before retrying or use cached data." | ||
| - pattern: "ETIMEDOUT|ECONNRESET|ECONNREFUSED" | ||
| hint: "Network error. Check connectivity and retry." | ||
| - pattern: "token.*expired|unauthorized|HTTP 401" | ||
| hint: "Authentication failure. Refresh credentials or tokens." | ||
| - pattern: "disk.*(full|space)|No space left" | ||
| hint: "Disk space exhausted. Free space before retrying." |
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
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,304 @@ | ||
| """Error classification and recovery hint system. | ||
|
|
||
| Classifies tool execution failures into a taxonomy aligned with ADR-035 | ||
| exit codes. Provides recovery hints from a YAML configuration file. | ||
| Logs errors for pattern learning and graduation to MEMORY.md. | ||
|
|
||
| Exit Codes (ADR-035): | ||
| 0 = success | ||
| 1 = logic/validation error | ||
| 2 = configuration error | ||
| 3 = external service error | ||
| 4 = authentication error | ||
| """ | ||
|
|
||
| from __future__ import annotations | ||
|
|
||
| import enum | ||
| import json | ||
| import logging | ||
| import re | ||
| from collections import Counter | ||
| from dataclasses import asdict, dataclass, field | ||
| from datetime import UTC, datetime | ||
| from pathlib import Path | ||
| from typing import Any | ||
|
|
||
| import yaml | ||
|
|
||
| _logger = logging.getLogger(__name__) | ||
|
|
||
| # Default paths for error logging and graduation. | ||
| _DEFAULT_ERROR_LOG = Path(".agents/sessions/errors.jsonl") | ||
| _GRADUATION_THRESHOLD = 3 # Patterns with 3+ successful recoveries graduate. | ||
|
|
||
|
|
||
| class ErrorType(enum.Enum): | ||
| """Error taxonomy aligned with ADR-035 exit codes.""" | ||
|
|
||
| TOOL_FAILURE = "tool_failure" | ||
| REASONING_DRIFT = "reasoning_drift" | ||
| INFINITE_LOOP = "infinite_loop" | ||
| SCOPE_CREEP = "scope_creep" | ||
| CONTEXT_OVERFLOW = "context_overflow" | ||
|
|
||
|
|
||
| # Map ADR-035 exit codes to error types where deterministic. | ||
| _EXIT_CODE_MAP: dict[int, ErrorType] = { | ||
| 2: ErrorType.TOOL_FAILURE, # config error | ||
| 3: ErrorType.TOOL_FAILURE, # external service error | ||
| 4: ErrorType.TOOL_FAILURE, # auth error | ||
| } | ||
|
|
||
| # Patterns indicating transient / retriable failures. | ||
| _TRANSIENT_PATTERNS: list[re.Pattern[str]] = [ | ||
| re.compile(r"rate limit", re.IGNORECASE), | ||
| re.compile(r"HTTP 429", re.IGNORECASE), | ||
| re.compile(r"HTTP 503", re.IGNORECASE), | ||
| re.compile(r"ETIMEDOUT", re.IGNORECASE), | ||
| re.compile(r"ECONNRESET", re.IGNORECASE), | ||
| ] | ||
|
|
||
|
|
||
| @dataclass(frozen=True) | ||
| class RecoveryHint: | ||
| """A single recovery hint for a failure pattern.""" | ||
|
|
||
| pattern: str | ||
| hint: str | ||
| compiled_pattern: re.Pattern[str] = field(init=False, repr=False, compare=False) | ||
|
|
||
| def __post_init__(self) -> None: | ||
| object.__setattr__( | ||
| self, | ||
| "compiled_pattern", | ||
| re.compile(self.pattern, re.IGNORECASE), | ||
| ) | ||
|
|
||
|
|
||
| @dataclass(frozen=True) | ||
| class ClassifiedError: | ||
| """Result of classifying a tool execution failure.""" | ||
|
|
||
| error_type: ErrorType | ||
| tool_name: str | ||
| exit_code: int | ||
| stderr: str | ||
| is_transient: bool | ||
| recovery_hints: tuple[str, ...] | ||
|
|
||
|
|
||
| def _is_transient(stderr: str) -> bool: | ||
| """Return True if stderr matches a known transient failure pattern.""" | ||
| return any(p.search(stderr) for p in _TRANSIENT_PATTERNS) | ||
|
|
||
|
|
||
| def load_recovery_hints( | ||
| hints_path: Path | None = None, | ||
| ) -> dict[str, list[RecoveryHint]]: | ||
| """Load recovery hints from YAML. | ||
|
|
||
| Args: | ||
| hints_path: Path to recovery-hints.yaml. Defaults to | ||
| .agents/recovery-hints.yaml relative to the repo root. | ||
|
|
||
| Returns: | ||
| Dict mapping section names to lists of RecoveryHint objects. | ||
| """ | ||
| if hints_path is None: | ||
| # Default path: scripts/ is at repo root, so parents[1] is repo root. | ||
| # Override via hints_path parameter for different layouts. | ||
| hints_path = Path(__file__).resolve().parents[1] / ".agents" / "recovery-hints.yaml" | ||
|
|
||
| if not hints_path.is_file(): | ||
| return {} | ||
|
|
||
| raw: dict[str, Any] = yaml.safe_load(hints_path.read_text(encoding="utf-8")) or {} | ||
| result: dict[str, list[RecoveryHint]] = {} | ||
|
|
||
| for section, entries in raw.items(): | ||
| if not isinstance(entries, list): | ||
| continue | ||
| hints: list[RecoveryHint] = [] | ||
| for entry in entries: | ||
| if isinstance(entry, dict) and "pattern" in entry and "hint" in entry: | ||
| hints.append(RecoveryHint(pattern=entry["pattern"], hint=entry["hint"])) | ||
rjmurillo-bot marked this conversation as resolved.
Show resolved
Hide resolved
|
||
| else: | ||
| _logger.warning("Skipped malformed entry in section '%s': %s", section, entry) | ||
| if hints: | ||
| result[section] = hints | ||
|
|
||
| return result | ||
|
|
||
|
|
||
| def _match_hints( | ||
| stderr: str, | ||
| tool_name: str, | ||
| hints_db: dict[str, list[RecoveryHint]], | ||
| ) -> tuple[str, ...]: | ||
| """Return matching recovery hints for the given failure.""" | ||
| matched: list[str] = [] | ||
|
|
||
| # Check tool-specific hints first. | ||
| tool_hints = hints_db.get(f"tool_{tool_name}", []) | ||
| for rh in tool_hints: | ||
| if rh.compiled_pattern.search(stderr): | ||
| matched.append(rh.hint) | ||
|
|
||
| # Check general hints. | ||
| for rh in hints_db.get("general", []): | ||
| if rh.compiled_pattern.search(stderr): | ||
| matched.append(rh.hint) | ||
|
|
||
| return tuple(matched) | ||
|
|
||
|
|
||
| def classify_error( | ||
| tool_name: str, | ||
| exit_code: int, | ||
| stderr: str, | ||
| *, | ||
| call_history: list[str] | None = None, | ||
| hints_db: dict[str, list[RecoveryHint]] | None = None, | ||
| ) -> ClassifiedError: | ||
| """Classify a tool execution failure. | ||
|
|
||
| Args: | ||
| tool_name: Name of the tool that failed. | ||
| exit_code: Process exit code. | ||
| stderr: Standard error output. | ||
| call_history: Recent tool call names for loop detection. | ||
| hints_db: Pre-loaded recovery hints. Loaded from disk if None. | ||
|
|
||
| Returns: | ||
| ClassifiedError with type, transient flag, and recovery hints. | ||
| """ | ||
| if hints_db is None: | ||
| hints_db = load_recovery_hints() | ||
|
|
||
| # Loop detection: 3+ consecutive identical calls. | ||
| if call_history and len(call_history) >= 3: | ||
| last_three = call_history[-3:] | ||
| if len(set(last_three)) == 1 and last_three[0] == tool_name: | ||
| return ClassifiedError( | ||
| error_type=ErrorType.INFINITE_LOOP, | ||
| tool_name=tool_name, | ||
| exit_code=exit_code, | ||
| stderr=stderr, | ||
| is_transient=False, | ||
| recovery_hints=( | ||
| "Loop detected: same tool called 3+ times. " | ||
| "Break the loop, summarize progress, try a different approach.", | ||
| ), | ||
| ) | ||
|
|
||
| # Exit code mapping. | ||
| error_type = _EXIT_CODE_MAP.get(exit_code, ErrorType.TOOL_FAILURE) | ||
|
|
||
| transient = _is_transient(stderr) | ||
| hints = _match_hints(stderr, tool_name, hints_db) | ||
|
|
||
| return ClassifiedError( | ||
| error_type=error_type, | ||
| tool_name=tool_name, | ||
| exit_code=exit_code, | ||
| stderr=stderr, | ||
| is_transient=transient, | ||
| recovery_hints=hints, | ||
| ) | ||
|
|
||
|
|
||
| @dataclass | ||
| class ErrorLogEntry: | ||
| """A single error log entry for pattern learning.""" | ||
|
|
||
| timestamp: str | ||
| error_type: str | ||
| tool: str | ||
| exit_code: int | ||
| recovery: str | ||
| success: bool | ||
|
|
||
|
|
||
| def log_error( | ||
| classified: ClassifiedError, | ||
| recovery_action: str, | ||
| success: bool, | ||
| *, | ||
| log_path: Path | None = None, | ||
| ) -> None: | ||
| """Log an error and its recovery outcome for pattern learning. | ||
|
|
||
| Args: | ||
| classified: The classified error from classify_error(). | ||
| recovery_action: Description of the recovery action taken. | ||
| success: True if the recovery was successful. | ||
| log_path: Path to errors.jsonl. Defaults to .agents/sessions/errors.jsonl. | ||
| """ | ||
| if log_path is None: | ||
| log_path = _DEFAULT_ERROR_LOG | ||
|
|
||
| log_path.parent.mkdir(parents=True, exist_ok=True) | ||
|
|
||
| entry = ErrorLogEntry( | ||
| timestamp=datetime.now(UTC).isoformat(), | ||
| error_type=classified.error_type.value, | ||
| tool=classified.tool_name, | ||
| exit_code=classified.exit_code, | ||
| recovery=recovery_action, | ||
| success=success, | ||
| ) | ||
|
|
||
| with log_path.open("a", encoding="utf-8") as f: | ||
| f.write(json.dumps(asdict(entry)) + "\n") | ||
|
|
||
|
|
||
| def get_graduation_candidates( | ||
| log_path: Path | None = None, | ||
| threshold: int = _GRADUATION_THRESHOLD, | ||
| ) -> list[dict[str, Any]]: | ||
| """Identify patterns eligible for graduation to MEMORY.md. | ||
|
|
||
| Patterns with `threshold` or more successful recoveries are candidates. | ||
|
|
||
| Args: | ||
| log_path: Path to errors.jsonl. Defaults to .agents/sessions/errors.jsonl. | ||
| threshold: Minimum successful recoveries required (default: 3). | ||
|
|
||
| Returns: | ||
| List of dicts with tool, error_type, recovery, and count. | ||
| """ | ||
| if log_path is None: | ||
| log_path = _DEFAULT_ERROR_LOG | ||
|
|
||
| if not log_path.is_file(): | ||
| return [] | ||
|
|
||
| success_counter: Counter[tuple[str, str, str]] = Counter() | ||
|
|
||
| with log_path.open("r", encoding="utf-8") as f: | ||
| for line in f: | ||
| line = line.strip() | ||
| if not line: | ||
| continue | ||
| try: | ||
| entry = json.loads(line) | ||
| if entry.get("success"): | ||
| key = (entry["tool"], entry["error_type"], entry["recovery"]) | ||
| success_counter[key] += 1 | ||
| except (json.JSONDecodeError, KeyError): | ||
| continue | ||
|
|
||
| candidates = [] | ||
| for (tool, error_type, recovery), count in success_counter.items(): | ||
| if count >= threshold: | ||
| candidates.append( | ||
| { | ||
| "tool": tool, | ||
| "error_type": error_type, | ||
| "recovery": recovery, | ||
| "count": count, | ||
| } | ||
| ) | ||
|
|
||
| return sorted(candidates, key=lambda x: x["count"], reverse=True) | ||
Oops, something went wrong.
Oops, something went wrong.
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.
Uh oh!
There was an error while loading. Please reload this page.