Files
automaker/apps/app/src/lib/log-parser.ts
Kacper 2a782392bc feat: enhance log parsing to support <summary> tags for structured output
- Introduced support for <summary> tags in log entries, allowing for better organization and parsing of summary content.
- Updated the detectEntryType function to recognize <summary> tags as a preferred format for summaries.
- Implemented summary accumulation logic to handle content between <summary> and </summary> tags.
- Modified the prompt in auto-mode service to instruct users to wrap their summaries in <summary> tags for consistency in log output.
2025-12-17 14:33:13 +01:00

744 lines
22 KiB
TypeScript

/**
* Log Parser Utility
* Parses agent output into structured sections for display
*/
export type LogEntryType =
| "prompt"
| "tool_call"
| "tool_result"
| "phase"
| "error"
| "success"
| "info"
| "debug"
| "warning"
| "thinking";
export type ToolCategory = 'read' | 'edit' | 'write' | 'bash' | 'search' | 'todo' | 'task' | 'other';
const TOOL_CATEGORIES: Record<string, ToolCategory> = {
'Read': 'read',
'Edit': 'edit',
'Write': 'write',
'Bash': 'bash',
'Grep': 'search',
'Glob': 'search',
'WebSearch': 'search',
'WebFetch': 'read',
'TodoWrite': 'todo',
'Task': 'task',
'NotebookEdit': 'edit',
'KillShell': 'bash',
};
/**
* Categorizes a tool name into a predefined category
*/
export function categorizeToolName(toolName: string): ToolCategory {
return TOOL_CATEGORIES[toolName] || 'other';
}
export interface LogEntryMetadata {
toolName?: string;
toolCategory?: ToolCategory;
filePath?: string;
summary?: string;
phase?: string;
}
export interface LogEntry {
id: string;
type: LogEntryType;
title: string;
content: string;
timestamp?: string;
collapsed?: boolean;
metadata?: LogEntryMetadata;
}
/**
* Generates a deterministic ID based on content and position
* This ensures the same log entry always gets the same ID,
* preserving expanded/collapsed state when new logs stream in
*
* Uses only the first 200 characters of content to ensure stability
* even when entries are merged (which appends content at the end)
*/
const generateDeterministicId = (content: string, lineIndex: number): string => {
// Use first 200 chars to ensure stability when entries are merged
const stableContent = content.slice(0, 200);
// Simple hash function for the content
let hash = 0;
const str = stableContent + '|' + lineIndex.toString();
for (let i = 0; i < str.length; i++) {
const char = str.charCodeAt(i);
hash = ((hash << 5) - hash) + char;
hash = hash & hash; // Convert to 32bit integer
}
return 'log_' + Math.abs(hash).toString(36);
};
/**
* Detects the type of log entry based on content patterns
*/
function detectEntryType(content: string): LogEntryType {
const trimmed = content.trim();
// Tool calls
if (trimmed.startsWith("🔧 Tool:") || trimmed.match(/^Tool:\s*/)) {
return "tool_call";
}
// Tool results / Input
if (trimmed.startsWith("Input:") || trimmed.startsWith("Result:") || trimmed.startsWith("Output:")) {
return "tool_result";
}
// Phase changes
if (
trimmed.startsWith("📋") ||
trimmed.startsWith("⚡") ||
trimmed.startsWith("✅") ||
trimmed.match(/^(Planning|Action|Verification)/i) ||
trimmed.match(/\[Phase:\s*([^\]]+)\]/) ||
trimmed.match(/Phase:\s*\w+/i)
) {
return "phase";
}
// Feature creation events
if (
trimmed.match(/\[Feature Creation\]/i) ||
trimmed.match(/Feature Creation/i) ||
trimmed.match(/Creating feature/i)
) {
return "success";
}
// Errors
if (trimmed.startsWith("❌") || trimmed.toLowerCase().includes("error:")) {
return "error";
}
// Success messages and summary sections
if (
trimmed.startsWith("✅") ||
trimmed.toLowerCase().includes("success") ||
trimmed.toLowerCase().includes("completed") ||
// Summary tags (preferred format from agent)
trimmed.startsWith("<summary>") ||
// Markdown summary headers (fallback)
trimmed.match(/^##\s+(Summary|Feature|Changes|Implementation)/i) ||
trimmed.match(/^(I've|I have) (successfully |now )?(completed|finished|implemented)/i)
) {
return "success";
}
// Warnings
if (trimmed.startsWith("⚠️") || trimmed.toLowerCase().includes("warning:")) {
return "warning";
}
// Thinking/Preparation info (be specific to avoid matching summary content)
if (
trimmed.toLowerCase().includes("ultrathink") ||
trimmed.match(/thinking level[:\s]*(low|medium|high|none|\d)/i) ||
trimmed.match(/^thinking level\s*$/i) ||
trimmed.toLowerCase().includes("estimated cost") ||
trimmed.toLowerCase().includes("estimated time") ||
trimmed.toLowerCase().includes("budget tokens") ||
trimmed.match(/thinking.*preparation/i)
) {
return "thinking";
}
// Debug info (JSON, stack traces, etc.)
if (
trimmed.startsWith("{") ||
trimmed.startsWith("[") ||
trimmed.includes("at ") ||
trimmed.match(/^\s*\d+\s*\|/)
) {
return "debug";
}
// Default to info
return "info";
}
/**
* Extracts tool name from a tool call entry
* Matches both "🔧 Tool: Name" and "Tool: Name" formats
*/
function extractToolName(content: string): string | undefined {
// Try emoji format first, then plain format
const match = content.match(/(?:🔧\s*)?Tool:\s*(\S+)/);
return match?.[1];
}
/**
* Extracts phase name from a phase entry
*/
function extractPhase(content: string): string | undefined {
if (content.includes("📋")) return "planning";
if (content.includes("⚡")) return "action";
if (content.includes("✅")) return "verification";
// Extract from [Phase: ...] format
const phaseMatch = content.match(/\[Phase:\s*([^\]]+)\]/);
if (phaseMatch) {
return phaseMatch[1].toLowerCase();
}
const match = content.match(/^(Planning|Action|Verification)/i);
return match?.[1]?.toLowerCase();
}
/**
* Extracts file path from tool input JSON
*/
function extractFilePath(content: string): string | undefined {
try {
const inputMatch = content.match(/Input:\s*([\s\S]*)/);
if (!inputMatch) return undefined;
const jsonStr = inputMatch[1].trim();
const parsed = JSON.parse(jsonStr) as Record<string, unknown>;
if (typeof parsed.file_path === 'string') return parsed.file_path;
if (typeof parsed.path === 'string') return parsed.path;
if (typeof parsed.notebook_path === 'string') return parsed.notebook_path;
return undefined;
} catch {
return undefined;
}
}
/**
* Generates a smart summary for tool calls based on the tool name and input
*/
export function generateToolSummary(toolName: string, content: string): string | undefined {
try {
// Try to parse JSON input
const inputMatch = content.match(/Input:\s*([\s\S]*)/);
if (!inputMatch) return undefined;
const jsonStr = inputMatch[1].trim();
const parsed = JSON.parse(jsonStr) as Record<string, unknown>;
switch (toolName) {
case 'Read': {
const filePath = parsed.file_path as string | undefined;
return `Reading ${filePath?.split('/').pop() || 'file'}`;
}
case 'Edit': {
const filePath = parsed.file_path as string | undefined;
const fileName = filePath?.split('/').pop() || 'file';
return `Editing ${fileName}`;
}
case 'Write': {
const filePath = parsed.file_path as string | undefined;
return `Writing ${filePath?.split('/').pop() || 'file'}`;
}
case 'Bash': {
const command = parsed.command as string | undefined;
const cmd = command?.slice(0, 50) || '';
return `Running: ${cmd}${(command?.length || 0) > 50 ? '...' : ''}`;
}
case 'Grep': {
const pattern = parsed.pattern as string | undefined;
return `Searching for "${pattern?.slice(0, 30) || ''}"`;
}
case 'Glob': {
const pattern = parsed.pattern as string | undefined;
return `Finding files: ${pattern || ''}`;
}
case 'TodoWrite': {
const todos = parsed.todos as unknown[] | undefined;
const todoCount = todos?.length || 0;
return `${todoCount} todo item${todoCount !== 1 ? 's' : ''}`;
}
case 'Task': {
const subagentType = parsed.subagent_type as string | undefined;
const description = parsed.description as string | undefined;
return `${subagentType || 'Agent'}: ${description || ''}`;
}
case 'WebSearch': {
const query = parsed.query as string | undefined;
return `Searching: "${query?.slice(0, 40) || ''}"`;
}
case 'WebFetch': {
const url = parsed.url as string | undefined;
return `Fetching: ${url?.slice(0, 40) || ''}`;
}
case 'NotebookEdit': {
const notebookPath = parsed.notebook_path as string | undefined;
return `Editing notebook: ${notebookPath?.split('/').pop() || 'notebook'}`;
}
case 'KillShell': {
return 'Terminating shell session';
}
default:
return undefined;
}
} catch {
return undefined;
}
}
/**
* Determines if an entry should be collapsed by default
*/
export function shouldCollapseByDefault(entry: LogEntry): boolean {
// Collapse if content is long
if (entry.content.length > 200) return true;
// Collapse if contains multi-line JSON (> 5 lines)
const lineCount = entry.content.split('\n').length;
if (lineCount > 5 && (entry.content.includes('{') || entry.content.includes('['))) {
return true;
}
// Collapse TodoWrite with multiple items
if (entry.metadata?.toolName === 'TodoWrite') {
try {
const inputMatch = entry.content.match(/Input:\s*([\s\S]*)/);
if (inputMatch) {
const parsed = JSON.parse(inputMatch[1].trim()) as Record<string, unknown>;
const todos = parsed.todos as unknown[] | undefined;
if (todos && todos.length > 1) return true;
}
} catch {
// Ignore parse errors
}
}
// Collapse Edit with code blocks
if (entry.metadata?.toolName === 'Edit' && entry.content.includes('old_string')) {
return true;
}
return false;
}
/**
* Generates a title for a log entry
*/
function generateTitle(type: LogEntryType, content: string): string {
switch (type) {
case "tool_call": {
const toolName = extractToolName(content);
return toolName ? `Tool Call: ${toolName}` : "Tool Call";
}
case "tool_result":
return "Tool Input/Result";
case "phase": {
const phase = extractPhase(content);
if (phase) {
// Capitalize first letter of each word
const formatted = phase.split(/\s+/).map(word =>
word.charAt(0).toUpperCase() + word.slice(1)
).join(" ");
return `Phase: ${formatted}`;
}
return "Phase Change";
}
case "error":
return "Error";
case "success": {
// Check if it's a summary section
if (content.startsWith("<summary>") || content.includes("<summary>")) {
return "Summary";
}
if (content.match(/^##\s+(Summary|Feature|Changes|Implementation)/i)) {
return "Summary";
}
if (content.match(/^All tasks completed/i) || content.match(/^(I've|I have) (successfully |now )?(completed|finished|implemented)/i)) {
return "Summary";
}
return "Success";
}
case "warning":
return "Warning";
case "thinking":
return "Thinking Level";
case "debug":
return "Debug Info";
case "prompt":
return "Prompt";
default:
return "Info";
}
}
/**
* Tracks bracket depth for JSON accumulation
*/
function calculateBracketDepth(line: string): { braceChange: number; bracketChange: number } {
let braceChange = 0;
let bracketChange = 0;
let inString = false;
let escapeNext = false;
for (const char of line) {
if (escapeNext) {
escapeNext = false;
continue;
}
if (char === '\\') {
escapeNext = true;
continue;
}
if (char === '"') {
inString = !inString;
continue;
}
if (inString) continue;
if (char === '{') braceChange++;
else if (char === '}') braceChange--;
else if (char === '[') bracketChange++;
else if (char === ']') bracketChange--;
}
return { braceChange, bracketChange };
}
/**
* Parses raw log output into structured entries
*/
export function parseLogOutput(rawOutput: string): LogEntry[] {
if (!rawOutput || !rawOutput.trim()) {
return [];
}
const entries: LogEntry[] = [];
const lines = rawOutput.split("\n");
let currentEntry: Omit<LogEntry, 'id'> & { id?: string } | null = null;
let currentContent: string[] = [];
let entryStartLine = 0; // Track the starting line for deterministic ID generation
// JSON accumulation state
let inJsonAccumulation = false;
let jsonBraceDepth = 0;
let jsonBracketDepth = 0;
// Summary tag accumulation state
let inSummaryAccumulation = false;
const finalizeEntry = () => {
if (currentEntry && currentContent.length > 0) {
currentEntry.content = currentContent.join("\n").trim();
if (currentEntry.content) {
// Populate enhanced metadata for tool calls
const toolName = currentEntry.metadata?.toolName;
if (toolName && currentEntry.type === 'tool_call') {
const toolCategory = categorizeToolName(toolName);
const filePath = extractFilePath(currentEntry.content);
const summary = generateToolSummary(toolName, currentEntry.content);
currentEntry.metadata = {
...currentEntry.metadata,
toolCategory,
filePath,
summary,
};
}
// Generate deterministic ID based on content and position
const entryWithId: LogEntry = {
...currentEntry as Omit<LogEntry, 'id'>,
id: generateDeterministicId(currentEntry.content, entryStartLine),
};
entries.push(entryWithId);
}
}
currentContent = [];
inJsonAccumulation = false;
jsonBraceDepth = 0;
jsonBracketDepth = 0;
inSummaryAccumulation = false;
};
let lineIndex = 0;
for (const line of lines) {
const trimmedLine = line.trim();
// Skip empty lines at the beginning
if (!trimmedLine && !currentEntry) {
lineIndex++;
continue;
}
// If we're in JSON accumulation mode, keep accumulating until depth returns to 0
if (inJsonAccumulation) {
currentContent.push(line);
const { braceChange, bracketChange } = calculateBracketDepth(trimmedLine);
jsonBraceDepth += braceChange;
jsonBracketDepth += bracketChange;
// JSON is complete when depth returns to 0
if (jsonBraceDepth <= 0 && jsonBracketDepth <= 0) {
inJsonAccumulation = false;
jsonBraceDepth = 0;
jsonBracketDepth = 0;
}
lineIndex++;
continue;
}
// If we're in summary accumulation mode, keep accumulating until </summary>
if (inSummaryAccumulation) {
currentContent.push(line);
// Summary is complete when we see closing tag
if (trimmedLine.includes("</summary>")) {
inSummaryAccumulation = false;
// Don't finalize here - let normal flow handle it
}
lineIndex++;
continue;
}
// Detect if this line starts a new entry
const lineType = detectEntryType(trimmedLine);
const isNewEntry =
trimmedLine.startsWith("🔧") ||
trimmedLine.startsWith("📋") ||
trimmedLine.startsWith("⚡") ||
trimmedLine.startsWith("✅") ||
trimmedLine.startsWith("❌") ||
trimmedLine.startsWith("⚠️") ||
trimmedLine.startsWith("🧠") ||
trimmedLine.match(/\[Phase:\s*([^\]]+)\]/) ||
trimmedLine.match(/\[Feature Creation\]/i) ||
trimmedLine.match(/\[Tool\]/i) ||
trimmedLine.match(/\[Agent\]/i) ||
trimmedLine.match(/\[Complete\]/i) ||
trimmedLine.match(/\[ERROR\]/i) ||
trimmedLine.match(/\[Status\]/i) ||
trimmedLine.toLowerCase().includes("ultrathink preparation") ||
trimmedLine.match(/thinking level[:\s]*(low|medium|high|none|\d)/i) ||
// Summary tags (preferred format from agent)
trimmedLine.startsWith("<summary>") ||
// Agent summary sections (markdown headers - fallback)
trimmedLine.match(/^##\s+(Summary|Feature|Changes|Implementation)/i) ||
// Summary introduction lines
trimmedLine.match(/^All tasks completed/i) ||
trimmedLine.match(/^(I've|I have) (successfully |now )?(completed|finished|implemented)/i);
// Check if this is an Input: line that should trigger JSON accumulation
const isInputLine = trimmedLine.startsWith("Input:") && currentEntry?.type === "tool_call";
if (isNewEntry) {
// Finalize previous entry
finalizeEntry();
// Track starting line for deterministic ID
entryStartLine = lineIndex;
// Start new entry (ID will be generated when finalizing)
currentEntry = {
type: lineType,
title: generateTitle(lineType, trimmedLine),
content: "",
metadata: {
toolName: extractToolName(trimmedLine),
phase: extractPhase(trimmedLine),
},
};
currentContent.push(trimmedLine);
// If this is a <summary> tag, start summary accumulation mode
if (trimmedLine.startsWith("<summary>") && !trimmedLine.includes("</summary>")) {
inSummaryAccumulation = true;
}
} else if (isInputLine && currentEntry) {
// Start JSON accumulation mode
currentContent.push(trimmedLine);
// Check if there's JSON on the same line after "Input:"
const inputContent = trimmedLine.replace(/^Input:\s*/, '');
if (inputContent) {
const { braceChange, bracketChange } = calculateBracketDepth(inputContent);
jsonBraceDepth = braceChange;
jsonBracketDepth = bracketChange;
// Only enter accumulation mode if JSON is incomplete
if (jsonBraceDepth > 0 || jsonBracketDepth > 0) {
inJsonAccumulation = true;
}
} else {
// Input: line with JSON starting on next line
inJsonAccumulation = true;
}
} else if (currentEntry) {
// Continue current entry
currentContent.push(line);
// Check if this line starts a JSON block
if (trimmedLine.startsWith('{') || trimmedLine.startsWith('[')) {
const { braceChange, bracketChange } = calculateBracketDepth(trimmedLine);
if (braceChange > 0 || bracketChange > 0) {
jsonBraceDepth = braceChange;
jsonBracketDepth = bracketChange;
if (jsonBraceDepth > 0 || jsonBracketDepth > 0) {
inJsonAccumulation = true;
}
}
}
} else {
// Track starting line for deterministic ID
entryStartLine = lineIndex;
// No current entry, create a default info entry
currentEntry = {
type: "info",
title: "Info",
content: "",
};
currentContent.push(line);
}
lineIndex++;
}
// Finalize last entry
finalizeEntry();
// Merge consecutive entries of the same type if they're both debug or info
const mergedEntries = mergeConsecutiveEntries(entries);
return mergedEntries;
}
/**
* Merges consecutive entries of the same type for cleaner display
*/
function mergeConsecutiveEntries(entries: LogEntry[]): LogEntry[] {
if (entries.length <= 1) return entries;
const merged: LogEntry[] = [];
let current: LogEntry | null = null;
let mergeIndex = 0;
for (const entry of entries) {
if (
current &&
(current.type === "debug" || current.type === "info") &&
current.type === entry.type
) {
// Merge into current - regenerate ID based on merged content
current.content += "\n\n" + entry.content;
current.id = generateDeterministicId(current.content, mergeIndex);
} else {
if (current) {
merged.push(current);
}
current = { ...entry };
mergeIndex = merged.length;
}
}
if (current) {
merged.push(current);
}
return merged;
}
/**
* Gets the color classes for a log entry type
*/
export function getLogTypeColors(type: LogEntryType): {
bg: string;
border: string;
text: string;
icon: string;
badge: string;
} {
switch (type) {
case "prompt":
return {
bg: "bg-blue-500/10",
border: "border-l-blue-500",
text: "text-blue-300",
icon: "text-blue-400",
badge: "bg-blue-500/20 text-blue-300",
};
case "tool_call":
return {
bg: "bg-amber-500/10",
border: "border-l-amber-500",
text: "text-amber-300",
icon: "text-amber-400",
badge: "bg-amber-500/20 text-amber-300",
};
case "tool_result":
return {
bg: "bg-slate-500/10",
border: "border-l-slate-400",
text: "text-slate-300",
icon: "text-slate-400",
badge: "bg-slate-500/20 text-slate-300",
};
case "phase":
return {
bg: "bg-cyan-500/10",
border: "border-l-cyan-500",
text: "text-cyan-300",
icon: "text-cyan-400",
badge: "bg-cyan-500/20 text-cyan-300",
};
case "error":
return {
bg: "bg-red-500/10",
border: "border-l-red-500",
text: "text-red-300",
icon: "text-red-400",
badge: "bg-red-500/20 text-red-300",
};
case "success":
return {
bg: "bg-emerald-500/10",
border: "border-l-emerald-500",
text: "text-emerald-300",
icon: "text-emerald-400",
badge: "bg-emerald-500/20 text-emerald-300",
};
case "warning":
return {
bg: "bg-orange-500/10",
border: "border-l-orange-500",
text: "text-orange-300",
icon: "text-orange-400",
badge: "bg-orange-500/20 text-orange-300",
};
case "thinking":
return {
bg: "bg-indigo-500/10",
border: "border-l-indigo-500",
text: "text-indigo-300",
icon: "text-indigo-400",
badge: "bg-indigo-500/20 text-indigo-300",
};
case "debug":
return {
bg: "bg-primary/10",
border: "border-l-primary",
text: "text-primary",
icon: "text-primary",
badge: "bg-primary/20 text-primary",
};
default:
return {
bg: "bg-zinc-500/10",
border: "border-l-zinc-500",
text: "text-zinc-300",
icon: "text-zinc-400",
badge: "bg-zinc-500/20 text-zinc-300",
};
}
}