How akhilkadarla/agentic_ai works
FinSight is a financial research assistant: users ask questions through the CLI or Streamlit UI, and a memory-enabled agent consults financial-data tools and Claude before returning research, optionally as a structured note. Separate branches cover SEC filings, Pillar 3 capital metrics and report search, MCP clients, and deployed access. The README documents these capabilities; some subsystem links rely on the source index because their implementation excerpts were not provided.
Main entry points: agent.py Research agent · cli.py CLI · tools.py Tool dispatch · ui.py Streamlit UI
Components 25
User interfaces
- CLI
src/finsight/cli.py - Streamlit UI
src/finsight/ui.py - Remote client
src/finsight/remote.py - FastAPI
src/finsight/api.py
Research workflow
- Research agent
src/finsight/agent.py - Tool dispatch
src/finsight/tools.py - Compliance checks
src/finsight/guardrail.py - Research notes
src/finsight/notes.py - Charts
src/finsight/charts.py
Financial data
- SEC EDGAR
src/finsight/sec.py - Filing extraction
src/finsight/filings.py - Knowledge Base
src/finsight/knowledge_base.py - Pillar 3 reports
src/finsight/pillar3.py - Capital metrics
src/finsight/data/pillar3_key_metrics.json
Model and runtime
- Tracing
src/finsight/tracing.py - Provider selection
src/finsight/providers.py - Claude calls
src/finsight/llm.py
Other components
- Analyst
- Anthropic / Bedrock
- SEC services
- Bedrock Knowledge Base
- Bedrock Guardrails
- AgentCore Runtime
- Cognito
- MCP client
Connections 29
- Analyst → CLI: asks via
README.md - Analyst → Streamlit UI: uses
README.md - CLI → Research agent: starts research
src/finsight/cli.py - Streamlit UI → Research agent: submits questions
src/finsight/ui.py - Streamlit UI → Charts: renders
src/finsight/ui.py - Research agent → Tool dispatch: dispatches tools
src/finsight/agent.py - Research agent → Compliance checks: checks answers
src/finsight/agent.py - Research agent → Claude calls: requests completions
src/finsight/agent.py - Research agent → Research notes: writes notes
src/finsight/agent.py - Research agent → Tracing: records traces
src/finsight/agent.py - Claude calls → Provider selection: selects client
src/finsight/llm.py - Provider selection → Anthropic / Bedrock: calls
src/finsight/providers.py - Tool dispatch → SEC EDGAR: fetches filings and facts
src/finsight/tools.py - SEC EDGAR → SEC services: requests data
src/finsight/sec.py - Tool dispatch → Knowledge Base: searches text
src/finsight/tools.py - Knowledge Base → Bedrock Knowledge Base: queries
README.md - Tool dispatch → Pillar 3 reports: gets metrics and passages
src/finsight/tools.py - Pillar 3 reports → Capital metrics: reads
README.md - Compliance checks → Bedrock Guardrails: applies policy
src/finsight/guardrail.py - Remote client → Cognito: authenticates
src/finsight/remote.py - Remote client → AgentCore Runtime: invokes
README.md - AgentCore Runtime → FastAPI: routes requests
README.md - FastAPI → Research agent: serves agent
src/finsight/api.py - MCP client → CLI: launches MCP
README.md - CLI → Tool dispatch: exposes research tools
README.md - Filing extraction → SEC EDGAR: uses EDGAR data
src/finsight/filings.py - CLI → Filing extraction: indexes filings
src/finsight/cli.py - CLI → Knowledge Base: manages indexing
src/finsight/cli.py - CLI → Pillar 3 reports: builds report metrics
src/finsight/cli.py