pankajbishnoi-tech/xpert

How pankajbishnoi-tech/xpert works

Cold Start turns a WhatsApp export into a structured personal map: people and circles, organisations and chat ties, goals, and open loops with suggested help. The main workflow loads and indexes chat data, runs domain-specific extraction, combines the results, and writes output.json; model calls are optional, with rules-based fallback. The README also documents output checking. Relationships follow the README and the shown source/index references; unsampled implementation details are not assumed.

Main entry points: solution.py Build and combine · data.py Load and index · llm.py LLM client · orgs.py Organisations and chats

Components 12

Data ingestion

Load and index xpert/data.py

Domain analysis

Organisations and chats xpert/orgs.py
People and circles xpert/people.py
Goals xpert/goals.py
Open loops and help xpert/loops.py

Orchestration and model access

Build and combine solution.py
LLM client xpert/llm.py

Output checking

Validate and score check.py

Other components

User
WhatsApp export
OpenAI-compatible model API
output.json

Connections 19

User → WhatsApp export: provides README.md
WhatsApp export → Load and index: loads README.md
Build and combine → Load and index: loads solution.py
Build and combine → Organisations and chats: runs solution.py
Build and combine → People and circles: runs solution.py
Build and combine → Goals: runs solution.py
Build and combine → Open loops and help: runs solution.py
Organisations and chats → Load and index: uses xpert/orgs.py
People and circles → Load and index: uses xpert/people.py
Goals → Load and index: uses xpert/goals.py
Open loops and help → Load and index: uses xpert/loops.py
Organisations and chats → LLM client: requests xpert/orgs.py
People and circles → LLM client: requests xpert/people.py
Goals → LLM client: requests xpert/goals.py
Open loops and help → LLM client: requests xpert/loops.py
LLM client → OpenAI-compatible model API: calls xpert/llm.py
Build and combine → output.json: writes README.md
User → Validate and score: runs README.md
Validate and score → output.json: checks README.md

Drawn by GitDiagram on November 9, 2026 from the repository's file tree, README and source files. Read as Markdown ·