How shreesha1-ux/spark-fnol-multiagent works
Spark FNOL is an insurance-claims triage product: an adjuster or claimant submits a statement and photos, the system assesses coverage, damage and consistency, then routes the claim for fast-track handling or human review. The Streamlit cockpit and FastAPI/static cockpit are distinct user-facing entry points. Core processing, deterministic decisioning, JSON-backed state and surveyor recommendations are represented; some feature relationships are documented in the README rather than visible in the sampled source. The graph omits uncertain internal call details where excerpts do not establish them.
Main entry points: agents_crew.py Agent Pipeline · state_manager.py State Manager · server.py Claims API · app.py Adjuster Cockpit
Components 19
Intake & Cockpits
- Adjuster Cockpit
app.py - Claims API
server.py - Cockpit Web UI
static/index.html
Claims Assessment
- Agent Pipeline
agents_crew.py - Statement Extraction
tools.py - Photo Appraisal
tools.py - Policy Verification
tools.py - Fraud & Weather Checks
tools.py
Decision & Oversight
- Tier Decision Gate
decision_gate.py - Surveyor Recommendation
agents_crew.py
Claim State
- Claim Data Models
schemas.py - State Manager
state_manager.py - Claims Store
data/db1_claims.json - Policy Records
data/db2_policies.json - Repair Cost Table
data/repair_costs.json - Runtime Configuration
config.py
External Services
- Google Gemini
- Open-Meteo
Other components
- Adjuster / Claimant
Connections 25
- Adjuster / Claimant → Adjuster Cockpit: submits claims
README.md - Adjuster / Claimant → Claims API: submits claims
README.md - Adjuster Cockpit → Agent Pipeline: advances claims
app.py - Adjuster Cockpit → State Manager: reads and updates
app.py - Claims API → Cockpit Web UI: serves UI
server.py - Claims API → Agent Pipeline: runs processing
server.py - Claims API → State Manager: reads and updates
server.py - Agent Pipeline → Statement Extraction: extracts facts
agents_crew.py - Agent Pipeline → Photo Appraisal: assesses photos
agents_crew.py - Agent Pipeline → Policy Verification: checks coverage
agents_crew.py - Agent Pipeline → Fraud & Weather Checks: checks consistency
agents_crew.py - Agent Pipeline → Tier Decision Gate: routes claims
agents_crew.py - Agent Pipeline → State Manager: persists stages
agents_crew.py - Agent Pipeline → Surveyor Recommendation: recommends surveyor
README.md - Tier Decision Gate → Claim Data Models: produces decision fields
decision_gate.py - State Manager → Claim Data Models: loads and saves
state_manager.py - State Manager → Claims Store: reads and writes
state_manager.py - State Manager → Policy Records: reads
state_manager.py - State Manager → Repair Cost Table: reads
state_manager.py - Policy Verification → Policy Records: verifies schedule
tools.py - Photo Appraisal → Repair Cost Table: estimates repairs
tools.py - Photo Appraisal → Google Gemini: analyzes images
README.md - Statement Extraction → Google Gemini: extracts narrative facts
README.md - Fraud & Weather Checks → Open-Meteo: checks historical weather
tools.py - Runtime Configuration → Google Gemini: configures models
config.py