mirror of
https://github.com/QwenLM/qwen-code.git
synced 2026-04-30 20:50:34 +00:00
feat(insight): refactor data processing and add qualitative insights structure
This commit is contained in:
DragonnZhang
parent
f1214c90ea
commit
5a7dcce196
4 changed files with 884 additions and 225 deletions
|
|
@ -17,116 +17,29 @@ import type {
|
|||
StreakData,
|
||||
SessionFacets,
|
||||
} from '../types/StaticInsightTypes.js';
|
||||
import type {
|
||||
QualitativeInsights,
|
||||
InsightImpressiveWorkflows,
|
||||
InsightProjectAreas,
|
||||
InsightFutureOpportunities,
|
||||
InsightFrictionPoints,
|
||||
InsightMemorableMoment,
|
||||
InsightImprovements,
|
||||
InsightInteractionStyle,
|
||||
InsightAtAGlance,
|
||||
} from '../types/QualitativeInsightTypes.js';
|
||||
|
||||
// Prompt content from prompt.txt
|
||||
const ANALYSIS_PROMPT = `Analyze this Qwen Code session and extract structured facets.
|
||||
|
||||
CRITICAL GUIDELINES:
|
||||
|
||||
1. **goal_categories**: Count ONLY what the USER explicitly asked for.
|
||||
- DO NOT count Qwen's autonomous codebase exploration
|
||||
- DO NOT count work Qwen decided to do on its own
|
||||
- ONLY count when user says "can you...", "please...", "I need...", "let's..."
|
||||
|
||||
2. **user_satisfaction_counts**: Base ONLY on explicit user signals.
|
||||
- "Yay!", "great!", "perfect!" → happy
|
||||
- "thanks", "looks good", "that works" → satisfied
|
||||
- "ok, now let's..." (continuing without complaint) → likely_satisfied
|
||||
- "that's not right", "try again" → dissatisfied
|
||||
- "this is broken", "I give up" → frustrated
|
||||
|
||||
3. **friction_counts**: Be specific about what went wrong.
|
||||
- misunderstood_request: Qwen interpreted incorrectly
|
||||
- wrong_approach: Right goal, wrong solution method
|
||||
- buggy_code: Code didn't work correctly
|
||||
- user_rejected_action: User said no/stop to a tool call
|
||||
- excessive_changes: Over-engineered or changed too much
|
||||
|
||||
4. If very short or just warmup, use warmup_minimal for goal_category`;
|
||||
|
||||
const INSIGHT_SCHEMA = {
|
||||
type: 'object',
|
||||
properties: {
|
||||
underlying_goal: {
|
||||
type: 'string',
|
||||
description: 'What the user fundamentally wanted to achieve',
|
||||
},
|
||||
goal_categories: {
|
||||
type: 'object',
|
||||
additionalProperties: { type: 'number' },
|
||||
},
|
||||
outcome: {
|
||||
type: 'string',
|
||||
enum: [
|
||||
'fully_achieved',
|
||||
'mostly_achieved',
|
||||
'partially_achieved',
|
||||
'not_achieved',
|
||||
'unclear_from_transcript',
|
||||
],
|
||||
},
|
||||
user_satisfaction_counts: {
|
||||
type: 'object',
|
||||
additionalProperties: { type: 'number' },
|
||||
},
|
||||
Qwen_helpfulness: {
|
||||
type: 'string',
|
||||
enum: [
|
||||
'unhelpful',
|
||||
'slightly_helpful',
|
||||
'moderately_helpful',
|
||||
'very_helpful',
|
||||
'essential',
|
||||
],
|
||||
},
|
||||
session_type: {
|
||||
type: 'string',
|
||||
enum: [
|
||||
'single_task',
|
||||
'multi_task',
|
||||
'iterative_refinement',
|
||||
'exploration',
|
||||
'quick_question',
|
||||
],
|
||||
},
|
||||
friction_counts: {
|
||||
type: 'object',
|
||||
additionalProperties: { type: 'number' },
|
||||
},
|
||||
friction_detail: {
|
||||
type: 'string',
|
||||
description: 'One sentence describing friction or empty',
|
||||
},
|
||||
primary_success: {
|
||||
type: 'string',
|
||||
enum: [
|
||||
'none',
|
||||
'fast_accurate_search',
|
||||
'correct_code_edits',
|
||||
'good_explanations',
|
||||
'proactive_help',
|
||||
'multi_file_changes',
|
||||
'good_debugging',
|
||||
],
|
||||
},
|
||||
brief_summary: {
|
||||
type: 'string',
|
||||
description: 'One sentence: what user wanted and whether they got it',
|
||||
},
|
||||
},
|
||||
required: [
|
||||
'underlying_goal',
|
||||
'goal_categories',
|
||||
'outcome',
|
||||
'user_satisfaction_counts',
|
||||
'Qwen_helpfulness',
|
||||
'session_type',
|
||||
'friction_counts',
|
||||
'friction_detail',
|
||||
'primary_success',
|
||||
'brief_summary',
|
||||
],
|
||||
};
|
||||
import {
|
||||
PROMPT_IMPRESSIVE_WORKFLOWS,
|
||||
PROMPT_PROJECT_AREAS,
|
||||
PROMPT_FUTURE_OPPORTUNITIES,
|
||||
PROMPT_FRICTION_POINTS,
|
||||
PROMPT_MEMORABLE_MOMENT,
|
||||
PROMPT_IMPROVEMENTS,
|
||||
PROMPT_INTERACTION_STYLE,
|
||||
PROMPT_AT_A_GLANCE,
|
||||
ANALYSIS_PROMPT,
|
||||
} from '../prompts/InsightPrompts.js';
|
||||
|
||||
export class DataProcessor {
|
||||
constructor(private config: Config) {}
|
||||
|
|
@ -159,8 +72,7 @@ export class DataProcessor {
|
|||
if ('text' in part && part.text) {
|
||||
output += `[Assistant]: ${part.text}\n`;
|
||||
} else if ('functionCall' in part) {
|
||||
// eslint-disable-next-line @typescript-eslint/no-explicit-any
|
||||
const call = (part as any).functionCall;
|
||||
const call = part.functionCall;
|
||||
if (call) {
|
||||
output += `[Tool: ${call.name}]\n`;
|
||||
}
|
||||
|
|
@ -178,6 +90,90 @@ export class DataProcessor {
|
|||
): Promise<SessionFacets | null> {
|
||||
if (records.length === 0) return null;
|
||||
|
||||
const INSIGHT_SCHEMA = {
|
||||
type: 'object',
|
||||
properties: {
|
||||
underlying_goal: {
|
||||
type: 'string',
|
||||
description: 'What the user fundamentally wanted to achieve',
|
||||
},
|
||||
goal_categories: {
|
||||
type: 'object',
|
||||
additionalProperties: { type: 'number' },
|
||||
},
|
||||
outcome: {
|
||||
type: 'string',
|
||||
enum: [
|
||||
'fully_achieved',
|
||||
'mostly_achieved',
|
||||
'partially_achieved',
|
||||
'not_achieved',
|
||||
'unclear_from_transcript',
|
||||
],
|
||||
},
|
||||
user_satisfaction_counts: {
|
||||
type: 'object',
|
||||
additionalProperties: { type: 'number' },
|
||||
},
|
||||
Qwen_helpfulness: {
|
||||
type: 'string',
|
||||
enum: [
|
||||
'unhelpful',
|
||||
'slightly_helpful',
|
||||
'moderately_helpful',
|
||||
'very_helpful',
|
||||
'essential',
|
||||
],
|
||||
},
|
||||
session_type: {
|
||||
type: 'string',
|
||||
enum: [
|
||||
'single_task',
|
||||
'multi_task',
|
||||
'iterative_refinement',
|
||||
'exploration',
|
||||
'quick_question',
|
||||
],
|
||||
},
|
||||
friction_counts: {
|
||||
type: 'object',
|
||||
additionalProperties: { type: 'number' },
|
||||
},
|
||||
friction_detail: {
|
||||
type: 'string',
|
||||
description: 'One sentence describing friction or empty',
|
||||
},
|
||||
primary_success: {
|
||||
type: 'string',
|
||||
enum: [
|
||||
'none',
|
||||
'fast_accurate_search',
|
||||
'correct_code_edits',
|
||||
'good_explanations',
|
||||
'proactive_help',
|
||||
'multi_file_changes',
|
||||
'good_debugging',
|
||||
],
|
||||
},
|
||||
brief_summary: {
|
||||
type: 'string',
|
||||
description: 'One sentence: what user wanted and whether they got it',
|
||||
},
|
||||
},
|
||||
required: [
|
||||
'underlying_goal',
|
||||
'goal_categories',
|
||||
'outcome',
|
||||
'user_satisfaction_counts',
|
||||
'Qwen_helpfulness',
|
||||
'session_type',
|
||||
'friction_counts',
|
||||
'friction_detail',
|
||||
'primary_success',
|
||||
'brief_summary',
|
||||
],
|
||||
};
|
||||
|
||||
const sessionText = this.formatRecordsForAnalysis(records);
|
||||
const prompt = `${ANALYSIS_PROMPT}\n\nSESSION:\n${sessionText}`;
|
||||
|
||||
|
|
@ -367,14 +363,389 @@ export class DataProcessor {
|
|||
baseDir: string,
|
||||
facetsOutputDir?: string,
|
||||
): Promise<InsightData> {
|
||||
// Initialize data structures
|
||||
const heatmap: HeatMapData = {};
|
||||
const tokenUsage: TokenUsageData = {};
|
||||
const activeHours: { [hour: number]: number } = {};
|
||||
const sessionStartTimes: { [sessionId: string]: Date } = {};
|
||||
const sessionEndTimes: { [sessionId: string]: Date } = {};
|
||||
const allChatFiles = await this.scanChatFiles(baseDir);
|
||||
|
||||
// Store all valid chat file paths for LLM analysis
|
||||
const [metrics, facets] = await Promise.all([
|
||||
this.generateMetrics(allChatFiles),
|
||||
this.generateFacets(allChatFiles, facetsOutputDir),
|
||||
]);
|
||||
|
||||
const qualitative = await this.generateQualitativeInsights(metrics, facets);
|
||||
|
||||
return {
|
||||
...metrics,
|
||||
qualitative,
|
||||
};
|
||||
}
|
||||
|
||||
private async generateQualitativeInsights(
|
||||
metrics: Omit<InsightData, 'facets' | 'qualitative'>,
|
||||
facets: SessionFacets[],
|
||||
): Promise<QualitativeInsights | undefined> {
|
||||
if (facets.length === 0) {
|
||||
return undefined;
|
||||
}
|
||||
|
||||
console.log('Generating qualitative insights...');
|
||||
|
||||
const commonData = this.prepareCommonPromptData(metrics, facets);
|
||||
|
||||
const generate = async <T>(
|
||||
promptTemplate: string,
|
||||
schema: Record<string, unknown>,
|
||||
): Promise<T> => {
|
||||
const prompt = `${promptTemplate}\n\n${commonData}`;
|
||||
try {
|
||||
const result = await this.config.getBaseLlmClient().generateJson({
|
||||
model: this.config.getModel(),
|
||||
contents: [{ role: 'user', parts: [{ text: prompt }] }],
|
||||
schema,
|
||||
abortSignal: AbortSignal.timeout(60000),
|
||||
});
|
||||
return result as T;
|
||||
} catch (error) {
|
||||
console.error('Failed to generate insight:', error);
|
||||
throw error;
|
||||
}
|
||||
};
|
||||
|
||||
// Schemas for each insight type
|
||||
// We define simplified schemas here to guide the LLM.
|
||||
// The types are already defined in QualitativeInsightTypes.ts
|
||||
|
||||
// 1. Impressive Workflows
|
||||
const schemaImpressiveWorkflows = {
|
||||
type: 'object',
|
||||
properties: {
|
||||
intro: { type: 'string' },
|
||||
impressive_workflows: {
|
||||
type: 'array',
|
||||
items: {
|
||||
type: 'object',
|
||||
properties: {
|
||||
title: { type: 'string' },
|
||||
description: { type: 'string' },
|
||||
},
|
||||
required: ['title', 'description'],
|
||||
},
|
||||
},
|
||||
},
|
||||
required: ['intro', 'impressive_workflows'],
|
||||
};
|
||||
|
||||
// 2. Project Areas
|
||||
const schemaProjectAreas = {
|
||||
type: 'object',
|
||||
properties: {
|
||||
areas: {
|
||||
type: 'array',
|
||||
items: {
|
||||
type: 'object',
|
||||
properties: {
|
||||
name: { type: 'string' },
|
||||
session_count: { type: 'number' },
|
||||
description: { type: 'string' },
|
||||
},
|
||||
required: ['name', 'session_count', 'description'],
|
||||
},
|
||||
},
|
||||
},
|
||||
required: ['areas'],
|
||||
};
|
||||
|
||||
// 3. Future Opportunities
|
||||
const schemaFutureOpportunities = {
|
||||
type: 'object',
|
||||
properties: {
|
||||
intro: { type: 'string' },
|
||||
opportunities: {
|
||||
type: 'array',
|
||||
items: {
|
||||
type: 'object',
|
||||
properties: {
|
||||
title: { type: 'string' },
|
||||
whats_possible: { type: 'string' },
|
||||
how_to_try: { type: 'string' },
|
||||
copyable_prompt: { type: 'string' },
|
||||
},
|
||||
required: [
|
||||
'title',
|
||||
'whats_possible',
|
||||
'how_to_try',
|
||||
'copyable_prompt',
|
||||
],
|
||||
},
|
||||
},
|
||||
},
|
||||
required: ['intro', 'opportunities'],
|
||||
};
|
||||
|
||||
// 4. Friction Points
|
||||
const schemaFrictionPoints = {
|
||||
type: 'object',
|
||||
properties: {
|
||||
intro: { type: 'string' },
|
||||
categories: {
|
||||
type: 'array',
|
||||
items: {
|
||||
type: 'object',
|
||||
properties: {
|
||||
category: { type: 'string' },
|
||||
description: { type: 'string' },
|
||||
examples: { type: 'array', items: { type: 'string' } },
|
||||
},
|
||||
required: ['category', 'description', 'examples'],
|
||||
},
|
||||
},
|
||||
},
|
||||
required: ['intro', 'categories'],
|
||||
};
|
||||
|
||||
// 5. Memorable Moment
|
||||
const schemaMemorableMoment = {
|
||||
type: 'object',
|
||||
properties: {
|
||||
headline: { type: 'string' },
|
||||
detail: { type: 'string' },
|
||||
},
|
||||
required: ['headline', 'detail'],
|
||||
};
|
||||
|
||||
// 6. Improvements
|
||||
const schemaImprovements = {
|
||||
type: 'object',
|
||||
properties: {
|
||||
Qwen_md_additions: {
|
||||
type: 'array',
|
||||
items: {
|
||||
type: 'object',
|
||||
properties: {
|
||||
addition: { type: 'string' },
|
||||
why: { type: 'string' },
|
||||
prompt_scaffold: { type: 'string' },
|
||||
},
|
||||
required: ['addition', 'why', 'prompt_scaffold'],
|
||||
},
|
||||
},
|
||||
features_to_try: {
|
||||
type: 'array',
|
||||
items: {
|
||||
type: 'object',
|
||||
properties: {
|
||||
feature: { type: 'string' },
|
||||
one_liner: { type: 'string' },
|
||||
why_for_you: { type: 'string' },
|
||||
example_code: { type: 'string' },
|
||||
},
|
||||
required: ['feature', 'one_liner', 'why_for_you', 'example_code'],
|
||||
},
|
||||
},
|
||||
usage_patterns: {
|
||||
type: 'array',
|
||||
items: {
|
||||
type: 'object',
|
||||
properties: {
|
||||
title: { type: 'string' },
|
||||
suggestion: { type: 'string' },
|
||||
detail: { type: 'string' },
|
||||
copyable_prompt: { type: 'string' },
|
||||
},
|
||||
required: ['title', 'suggestion', 'detail', 'copyable_prompt'],
|
||||
},
|
||||
},
|
||||
},
|
||||
required: ['Qwen_md_additions', 'features_to_try', 'usage_patterns'],
|
||||
};
|
||||
|
||||
// 7. Interaction Style
|
||||
const schemaInteractionStyle = {
|
||||
type: 'object',
|
||||
properties: {
|
||||
narrative: { type: 'string' },
|
||||
key_pattern: { type: 'string' },
|
||||
},
|
||||
required: ['narrative', 'key_pattern'],
|
||||
};
|
||||
|
||||
// 8. At A Glance
|
||||
const schemaAtAGlance = {
|
||||
type: 'object',
|
||||
properties: {
|
||||
whats_working: { type: 'string' },
|
||||
whats_hindering: { type: 'string' },
|
||||
quick_wins: { type: 'string' },
|
||||
ambitious_workflows: { type: 'string' },
|
||||
},
|
||||
required: [
|
||||
'whats_working',
|
||||
'whats_hindering',
|
||||
'quick_wins',
|
||||
'ambitious_workflows',
|
||||
],
|
||||
};
|
||||
|
||||
const limit = pLimit(4);
|
||||
|
||||
try {
|
||||
const [
|
||||
impressiveWorkflows,
|
||||
projectAreas,
|
||||
futureOpportunities,
|
||||
frictionPoints,
|
||||
memorableMoment,
|
||||
improvements,
|
||||
interactionStyle,
|
||||
atAGlance,
|
||||
] = await Promise.all([
|
||||
limit(() =>
|
||||
generate<InsightImpressiveWorkflows>(
|
||||
PROMPT_IMPRESSIVE_WORKFLOWS,
|
||||
schemaImpressiveWorkflows,
|
||||
),
|
||||
),
|
||||
limit(() =>
|
||||
generate<InsightProjectAreas>(
|
||||
PROMPT_PROJECT_AREAS,
|
||||
schemaProjectAreas,
|
||||
),
|
||||
),
|
||||
limit(() =>
|
||||
generate<InsightFutureOpportunities>(
|
||||
PROMPT_FUTURE_OPPORTUNITIES,
|
||||
schemaFutureOpportunities,
|
||||
),
|
||||
),
|
||||
limit(() =>
|
||||
generate<InsightFrictionPoints>(
|
||||
PROMPT_FRICTION_POINTS,
|
||||
schemaFrictionPoints,
|
||||
),
|
||||
),
|
||||
limit(() =>
|
||||
generate<InsightMemorableMoment>(
|
||||
PROMPT_MEMORABLE_MOMENT,
|
||||
schemaMemorableMoment,
|
||||
),
|
||||
),
|
||||
limit(() =>
|
||||
generate<InsightImprovements>(
|
||||
PROMPT_IMPROVEMENTS,
|
||||
schemaImprovements,
|
||||
),
|
||||
),
|
||||
limit(() =>
|
||||
generate<InsightInteractionStyle>(
|
||||
PROMPT_INTERACTION_STYLE,
|
||||
schemaInteractionStyle,
|
||||
),
|
||||
),
|
||||
limit(() =>
|
||||
generate<InsightAtAGlance>(PROMPT_AT_A_GLANCE, schemaAtAGlance),
|
||||
),
|
||||
]);
|
||||
|
||||
return {
|
||||
impressiveWorkflows,
|
||||
projectAreas,
|
||||
futureOpportunities,
|
||||
frictionPoints,
|
||||
memorableMoment,
|
||||
improvements,
|
||||
interactionStyle,
|
||||
atAGlance,
|
||||
};
|
||||
} catch (e) {
|
||||
console.error('Error generating qualitative insights:', e);
|
||||
return undefined;
|
||||
}
|
||||
}
|
||||
|
||||
private prepareCommonPromptData(
|
||||
metrics: Omit<InsightData, 'facets' | 'qualitative'>,
|
||||
facets: SessionFacets[],
|
||||
): string {
|
||||
// 1. DATA section
|
||||
const goalsAgg: Record<string, number> = {};
|
||||
const outcomesAgg: Record<string, number> = {};
|
||||
const satisfactionAgg: Record<string, number> = {};
|
||||
const frictionAgg: Record<string, number> = {};
|
||||
const successAgg: Record<string, number> = {};
|
||||
|
||||
facets.forEach((facet) => {
|
||||
// Aggregate goals
|
||||
Object.entries(facet.goal_categories).forEach(([goal, count]) => {
|
||||
goalsAgg[goal] = (goalsAgg[goal] || 0) + count;
|
||||
});
|
||||
|
||||
// Aggregate outcomes
|
||||
outcomesAgg[facet.outcome] = (outcomesAgg[facet.outcome] || 0) + 1;
|
||||
|
||||
// Aggregate satisfaction
|
||||
Object.entries(facet.user_satisfaction_counts).forEach(([sat, count]) => {
|
||||
satisfactionAgg[sat] = (satisfactionAgg[sat] || 0) + count;
|
||||
});
|
||||
|
||||
// Aggregate friction
|
||||
Object.entries(facet.friction_counts).forEach(([fric, count]) => {
|
||||
frictionAgg[fric] = (frictionAgg[fric] || 0) + count;
|
||||
});
|
||||
|
||||
// Aggregate success (primary_success)
|
||||
if (facet.primary_success && facet.primary_success !== 'none') {
|
||||
successAgg[facet.primary_success] =
|
||||
(successAgg[facet.primary_success] || 0) + 1;
|
||||
}
|
||||
});
|
||||
|
||||
const topGoals = Object.entries(goalsAgg)
|
||||
.sort((a, b) => b[1] - a[1])
|
||||
.slice(0, 8);
|
||||
|
||||
const dataObj = {
|
||||
sessions: metrics.totalSessions || facets.length,
|
||||
analyzed: facets.length,
|
||||
date_range: {
|
||||
start: Object.keys(metrics.heatmap).sort()[0] || 'N/A',
|
||||
end: Object.keys(metrics.heatmap).sort().pop() || 'N/A',
|
||||
},
|
||||
messages: metrics.totalMessages || 0,
|
||||
hours: metrics.totalHours || 0,
|
||||
commits: 0, // Not tracked yet
|
||||
top_tools: metrics.topTools || [],
|
||||
top_goals: topGoals,
|
||||
outcomes: outcomesAgg,
|
||||
satisfaction: satisfactionAgg,
|
||||
friction: frictionAgg,
|
||||
success: successAgg,
|
||||
};
|
||||
|
||||
// 2. SESSION SUMMARIES section
|
||||
const sessionSummaries = facets
|
||||
.map((f) => `- ${f.brief_summary}`)
|
||||
.join('\n');
|
||||
|
||||
// 3. FRICTION DETAILS section
|
||||
const frictionDetails = facets
|
||||
.filter((f) => f.friction_detail && f.friction_detail.trim().length > 0)
|
||||
.map((f) => `- ${f.friction_detail}`)
|
||||
.join('\n');
|
||||
|
||||
return `DATA:
|
||||
${JSON.stringify(dataObj, null, 2)}
|
||||
|
||||
SESSION SUMMARIES:
|
||||
${sessionSummaries}
|
||||
|
||||
FRICTION DETAILS:
|
||||
${frictionDetails}
|
||||
|
||||
USER INSTRUCTIONS TO Qwen:
|
||||
None captured`;
|
||||
}
|
||||
|
||||
private async scanChatFiles(
|
||||
baseDir: string,
|
||||
): Promise<Array<{ path: string; mtime: number }>> {
|
||||
const allChatFiles: Array<{ path: string; mtime: number }> = [];
|
||||
|
||||
try {
|
||||
|
|
@ -390,11 +761,22 @@ export class DataProcessor {
|
|||
if (stats.isDirectory()) {
|
||||
const chatsDir = path.join(projectPath, 'chats');
|
||||
|
||||
let chatFiles: string[] = [];
|
||||
try {
|
||||
// Get all chat files in the chats directory
|
||||
const files = await fs.readdir(chatsDir);
|
||||
chatFiles = files.filter((file) => file.endsWith('.jsonl'));
|
||||
const chatFiles = files.filter((file) => file.endsWith('.jsonl'));
|
||||
|
||||
for (const file of chatFiles) {
|
||||
const filePath = path.join(chatsDir, file);
|
||||
|
||||
// Get file stats for sorting by recency
|
||||
try {
|
||||
const fileStats = await fs.stat(filePath);
|
||||
allChatFiles.push({ path: filePath, mtime: fileStats.mtimeMs });
|
||||
} catch (e) {
|
||||
console.error(`Failed to stat file ${filePath}:`, e);
|
||||
}
|
||||
}
|
||||
} catch (error) {
|
||||
if ((error as NodeJS.ErrnoException).code !== 'ENOENT') {
|
||||
console.log(
|
||||
|
|
@ -404,68 +786,157 @@ export class DataProcessor {
|
|||
// Continue to next project if chats directory doesn't exist
|
||||
continue;
|
||||
}
|
||||
|
||||
// Process each chat file in this project
|
||||
for (const file of chatFiles) {
|
||||
const filePath = path.join(chatsDir, file);
|
||||
|
||||
// Get file stats for sorting by recency
|
||||
try {
|
||||
const fileStats = await fs.stat(filePath);
|
||||
allChatFiles.push({ path: filePath, mtime: fileStats.mtimeMs });
|
||||
} catch (e) {
|
||||
console.error(`Failed to stat file ${filePath}:`, e);
|
||||
}
|
||||
|
||||
const records = await readJsonlFile<ChatRecord>(filePath);
|
||||
|
||||
// Process each record
|
||||
for (const record of records) {
|
||||
const timestamp = new Date(record.timestamp);
|
||||
const dateKey = this.formatDate(timestamp);
|
||||
const hour = timestamp.getHours();
|
||||
|
||||
// Update heatmap (count of interactions per day)
|
||||
heatmap[dateKey] = (heatmap[dateKey] || 0) + 1;
|
||||
|
||||
// Update active hours
|
||||
activeHours[hour] = (activeHours[hour] || 0) + 1;
|
||||
|
||||
// Update token usage
|
||||
if (record.usageMetadata) {
|
||||
const usage = tokenUsage[dateKey] || {
|
||||
input: 0,
|
||||
output: 0,
|
||||
total: 0,
|
||||
};
|
||||
|
||||
usage.input += record.usageMetadata.promptTokenCount || 0;
|
||||
usage.output += record.usageMetadata.candidatesTokenCount || 0;
|
||||
usage.total += record.usageMetadata.totalTokenCount || 0;
|
||||
|
||||
tokenUsage[dateKey] = usage;
|
||||
}
|
||||
|
||||
// Track session times
|
||||
if (!sessionStartTimes[record.sessionId]) {
|
||||
sessionStartTimes[record.sessionId] = timestamp;
|
||||
}
|
||||
sessionEndTimes[record.sessionId] = timestamp;
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
} catch (error) {
|
||||
if ((error as NodeJS.ErrnoException).code === 'ENOENT') {
|
||||
// Base directory doesn't exist, return empty insights
|
||||
// Base directory doesn't exist, return empty
|
||||
console.log(`Base directory does not exist: ${baseDir}`);
|
||||
} else {
|
||||
console.log(`Error reading base directory: ${error}`);
|
||||
}
|
||||
}
|
||||
|
||||
return allChatFiles;
|
||||
}
|
||||
|
||||
private async generateMetrics(
|
||||
files: Array<{ path: string; mtime: number }>,
|
||||
): Promise<Omit<InsightData, 'facets' | 'qualitative'>> {
|
||||
// Initialize data structures
|
||||
const heatmap: HeatMapData = {};
|
||||
const tokenUsage: TokenUsageData = {};
|
||||
const activeHours: { [hour: number]: number } = {};
|
||||
const sessionStartTimes: { [sessionId: string]: Date } = {};
|
||||
const sessionEndTimes: { [sessionId: string]: Date } = {};
|
||||
let totalMessages = 0;
|
||||
const toolUsage: Record<string, number> = {};
|
||||
|
||||
for (const fileInfo of files) {
|
||||
const records = await readJsonlFile<ChatRecord>(fileInfo.path);
|
||||
totalMessages += records.length;
|
||||
|
||||
// Process each record
|
||||
for (const record of records) {
|
||||
const timestamp = new Date(record.timestamp);
|
||||
const dateKey = this.formatDate(timestamp);
|
||||
const hour = timestamp.getHours();
|
||||
|
||||
// Update heatmap (count of interactions per day)
|
||||
heatmap[dateKey] = (heatmap[dateKey] || 0) + 1;
|
||||
|
||||
// Update active hours
|
||||
activeHours[hour] = (activeHours[hour] || 0) + 1;
|
||||
|
||||
// Update token usage
|
||||
if (record.usageMetadata) {
|
||||
const usage = tokenUsage[dateKey] || {
|
||||
input: 0,
|
||||
output: 0,
|
||||
total: 0,
|
||||
};
|
||||
|
||||
usage.input += record.usageMetadata.promptTokenCount || 0;
|
||||
usage.output += record.usageMetadata.candidatesTokenCount || 0;
|
||||
usage.total += record.usageMetadata.totalTokenCount || 0;
|
||||
|
||||
tokenUsage[dateKey] = usage;
|
||||
}
|
||||
|
||||
// Track session times
|
||||
if (!sessionStartTimes[record.sessionId]) {
|
||||
sessionStartTimes[record.sessionId] = timestamp;
|
||||
}
|
||||
sessionEndTimes[record.sessionId] = timestamp;
|
||||
|
||||
// Track tool usage
|
||||
if (record.type === 'assistant' && record.message?.parts) {
|
||||
for (const part of record.message.parts) {
|
||||
if ('functionCall' in part) {
|
||||
const name = part.functionCall!.name!;
|
||||
toolUsage[name] = (toolUsage[name] || 0) + 1;
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
// Calculate streak data
|
||||
const streakData = this.calculateStreaks(Object.keys(heatmap));
|
||||
|
||||
// Calculate longest work session and total hours
|
||||
let longestWorkDuration = 0;
|
||||
let longestWorkDate: string | null = null;
|
||||
let totalDurationMs = 0;
|
||||
|
||||
const sessionIds = Object.keys(sessionStartTimes);
|
||||
const totalSessions = sessionIds.length;
|
||||
|
||||
for (const sessionId of sessionIds) {
|
||||
const start = sessionStartTimes[sessionId];
|
||||
const end = sessionEndTimes[sessionId];
|
||||
const durationMs = end.getTime() - start.getTime();
|
||||
const durationMinutes = Math.round(durationMs / (1000 * 60));
|
||||
|
||||
totalDurationMs += durationMs;
|
||||
|
||||
if (durationMinutes > longestWorkDuration) {
|
||||
longestWorkDuration = durationMinutes;
|
||||
longestWorkDate = this.formatDate(start);
|
||||
}
|
||||
}
|
||||
|
||||
const totalHours = Math.round(totalDurationMs / (1000 * 60 * 60));
|
||||
|
||||
// Calculate latest active time
|
||||
let latestActiveTime: string | null = null;
|
||||
let latestTimestamp = new Date(0);
|
||||
for (const dateStr in heatmap) {
|
||||
const date = new Date(dateStr);
|
||||
if (date > latestTimestamp) {
|
||||
latestTimestamp = date;
|
||||
latestActiveTime = date.toLocaleTimeString([], {
|
||||
hour: '2-digit',
|
||||
minute: '2-digit',
|
||||
});
|
||||
}
|
||||
}
|
||||
|
||||
// Calculate top tools
|
||||
const topTools = Object.entries(toolUsage)
|
||||
.sort((a, b) => b[1] - a[1])
|
||||
.slice(0, 10);
|
||||
|
||||
// Calculate achievements
|
||||
const achievements = this.calculateAchievements(
|
||||
activeHours,
|
||||
heatmap,
|
||||
tokenUsage,
|
||||
);
|
||||
|
||||
return {
|
||||
heatmap,
|
||||
tokenUsage,
|
||||
currentStreak: streakData.currentStreak,
|
||||
longestStreak: streakData.longestStreak,
|
||||
longestWorkDate,
|
||||
longestWorkDuration,
|
||||
activeHours,
|
||||
latestActiveTime,
|
||||
achievements,
|
||||
totalSessions,
|
||||
totalMessages,
|
||||
totalHours,
|
||||
topTools,
|
||||
};
|
||||
}
|
||||
|
||||
private async generateFacets(
|
||||
allFiles: Array<{ path: string; mtime: number }>,
|
||||
facetsOutputDir?: string,
|
||||
): Promise<SessionFacets[]> {
|
||||
// Sort files by recency (descending) and take top 50
|
||||
const recentFiles = allChatFiles
|
||||
const recentFiles = [...allFiles]
|
||||
.sort((a, b) => b.mtime - a.mtime)
|
||||
.slice(0, 50);
|
||||
|
||||
|
|
@ -541,58 +1012,6 @@ export class DataProcessor {
|
|||
const facets = sessionFacetsWithNulls.filter(
|
||||
(f): f is SessionFacets => f !== null,
|
||||
);
|
||||
|
||||
// Calculate streak data
|
||||
const streakData = this.calculateStreaks(Object.keys(heatmap));
|
||||
|
||||
// Calculate longest work session
|
||||
let longestWorkDuration = 0;
|
||||
let longestWorkDate: string | null = null;
|
||||
for (const sessionId in sessionStartTimes) {
|
||||
const start = sessionStartTimes[sessionId];
|
||||
const end = sessionEndTimes[sessionId];
|
||||
const durationMinutes = Math.round(
|
||||
(end.getTime() - start.getTime()) / (1000 * 60),
|
||||
);
|
||||
|
||||
if (durationMinutes > longestWorkDuration) {
|
||||
longestWorkDuration = durationMinutes;
|
||||
longestWorkDate = this.formatDate(start);
|
||||
}
|
||||
}
|
||||
|
||||
// Calculate latest active time
|
||||
let latestActiveTime: string | null = null;
|
||||
let latestTimestamp = new Date(0);
|
||||
for (const dateStr in heatmap) {
|
||||
const date = new Date(dateStr);
|
||||
if (date > latestTimestamp) {
|
||||
latestTimestamp = date;
|
||||
latestActiveTime = date.toLocaleTimeString([], {
|
||||
hour: '2-digit',
|
||||
minute: '2-digit',
|
||||
});
|
||||
}
|
||||
}
|
||||
|
||||
// Calculate achievements
|
||||
const achievements = this.calculateAchievements(
|
||||
activeHours,
|
||||
heatmap,
|
||||
tokenUsage,
|
||||
);
|
||||
|
||||
return {
|
||||
heatmap,
|
||||
tokenUsage,
|
||||
currentStreak: streakData.currentStreak,
|
||||
longestStreak: streakData.longestStreak,
|
||||
longestWorkDate,
|
||||
longestWorkDuration,
|
||||
activeHours,
|
||||
latestActiveTime,
|
||||
achievements,
|
||||
facets,
|
||||
};
|
||||
return facets;
|
||||
}
|
||||
}
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue