How fedirrohalin/obrio-aiops-assignment works
OBRIO Support Automation has two runtime products: an AI ticket classifier for triage and a CLI assistant for support agents. The assistant accepts ticket text, loads and searches the knowledge base, then drafts a summary and tone-specific replies with a validated citation. The classifier turns ticket text into category, priority, and recommended action, with a fallback for failures. The graph reflects the shown assistant orchestration and classifier calls; classifier entry-point details and some runtime wiring are only documented or indexed, so uncertain links are omitted.
Main entry points: cli.py Assistant CLI · generator.py Reply generator · classifier.py Ticket triage · llm_client.py Assistant LLM client
Components 19
Ticket classifier
- Classifier entry point
src/main.py - Ticket triage
src/classifier.py - Triage prompts
src/prompts.py - Triage result
src/schemas.py
Agent assistant
- Assistant CLI
src/assistant/cli.py - Knowledge base loader
src/assistant/kb.py - Article retrieval
src/assistant/retriever.py - Reply prompts
src/assistant/prompts.py - Reply generator
src/assistant/generator.py - Retry and fallback
src/assistant/resilience.py - Assistant output
src/assistant/schemas.py
Models and configuration
- Assistant settings
src/assistant/config.py - Assistant LLM client
src/assistant/llm_client.py - Classifier LLM client
src/llm_client.py - Classifier configuration
src/config.py
Data and external services
- KB articles
data/kb/articles.json - OpenAI API
Other components
- Support agent
- Incoming ticket
Connections 22
- Support agent → Assistant CLI: submits ticket
README.md - Incoming ticket → Classifier entry point: provides ticket
README.md - Classifier entry point → Ticket triage: routes to
src/main.py - Ticket triage → Triage prompts: builds prompt
src/classifier.py - Ticket triage → Classifier LLM client: requests completion
src/classifier.py - Ticket triage → Triage result: validates output
src/classifier.py - Classifier LLM client → Classifier configuration: uses model
src/llm_client.py - Classifier LLM client → OpenAI API: calls API
src/llm_client.py - Assistant CLI → Assistant settings: loads settings
src/assistant/cli.py - Assistant CLI → Knowledge base loader: loads articles
src/assistant/cli.py - Knowledge base loader → KB articles: reads articles
src/assistant/kb.py - Assistant CLI → Article retrieval: searches articles
src/assistant/cli.py - Article retrieval → Knowledge base loader: gets articles
src/assistant/retriever.py - Assistant CLI → Assistant LLM client: initializes client
src/assistant/cli.py - Assistant CLI → Reply generator: requests drafts
src/assistant/cli.py - Reply generator → Reply prompts: builds prompt
src/assistant/generator.py - Reply generator → Assistant LLM client: requests completion
src/assistant/generator.py - Reply generator → Retry and fallback: uses retries
src/assistant/generator.py - Reply generator → Assistant output: validates output
src/assistant/generator.py - Assistant LLM client → Assistant settings: uses settings
src/assistant/llm_client.py - Assistant LLM client → OpenAI API: calls API
src/assistant/llm_client.py - Assistant CLI → Support agent: prints drafts
src/assistant/cli.py