# dfe-digital/document-analysis-agentic-poc architecture

> Architecture diagram of the GitHub repository dfe-digital/document-analysis-agentic-poc, generated by GitDiagram from its file tree, README and a sample of source files (last updated 2026-11-18). It is an AI-made overview: check details against the source.

- Interactive diagram: https://gitdiagram.com/dfe-digital/document-analysis-agentic-poc
- Repository: https://github.com/dfe-digital/document-analysis-agentic-poc

## Overview

This is a terminal-first, multi-agent document analysis proof of concept. Users add DOCX, PDF, or HTML documents to a persistent knowledge base, then search it or analyse a document or topic for duplicates, conflicts, and inconsistencies; results include summaries and agent metrics. The graph preserves both the programmatic detector path and the documented AutoGen orchestration/chat capabilities. Some wiring is supported by the README or SOURCE INDEX rather than sampled call sites, so those links cite that evidence; unsampled implementation details are not inferred.

Main entry points: [`src/agents/conflict_detector_agent.py`](https://github.com/dfe-digital/document-analysis-agentic-poc/blob/main/src/agents/conflict_detector_agent.py) (Analysis Facade), [`src/knowledge/vector_store.py`](https://github.com/dfe-digital/document-analysis-agentic-poc/blob/main/src/knowledge/vector_store.py) (Vector Store), [`src/agents/specialized/orchestrator_agent.py`](https://github.com/dfe-digital/document-analysis-agentic-poc/blob/main/src/agents/specialized/orchestrator_agent.py) (Orchestrator Agent), [`src/analysis/conflict_detector.py`](https://github.com/dfe-digital/document-analysis-agentic-poc/blob/main/src/analysis/conflict_detector.py) (Conflict Detector)

## Diagram (Mermaid)

```mermaid
flowchart TD

subgraph group_entry["User Entry"]
  node_cli["CLI Commands<br/>[main.py]"]
  node_workflow["Analysis Facade"]
end

subgraph group_knowledge["Knowledge Base"]
  node_loader["Document Loader<br/>[document_loader.py]"]
  node_store[("Vector Store<br/>[vector_store.py]")]
end

subgraph group_analysis["Analysis"]
  node_dupdet["Duplication Detector"]
  node_confdet["Conflict Detector"]
  node_incdet["Inconsistency Detector"]
  node_results["Analysis Models<br/>[models.py]"]
  node_summaries["Finding Summaries<br/>[one_liners.py]"]
end

subgraph group_agents["Agent Workflow"]
  node_orchestrator["Orchestrator Agent"]
  node_dupagent["Duplication Agent"]
  node_confagent["Conflict Agent<br/>[conflict_agent.py]"]
  node_incagent["Inconsistency Agent"]
  node_prompts["Agent Prompts<br/>[prompts.py]"]
end

node_user(("Terminal User"))
node_openai{{"OpenAI Models"}}

node_user -->|"runs commands"| node_cli
node_cli -->|"invokes analysis"| node_workflow
node_cli -->|"adds and searches"| node_workflow
node_workflow -->|"loads documents"| node_loader
node_workflow -->|"stores and retrieves"| node_store
node_loader -->|"provides chunks"| node_store
node_workflow -->|"runs detection"| node_dupdet
node_workflow -->|"runs detection"| node_confdet
node_workflow -->|"runs detection"| node_incdet
node_dupdet -->|"searches similar chunks"| node_store
node_confdet -->|"retrieves candidates"| node_store
node_incdet -->|"uses semantic search"| node_store
node_confdet -->|"analyses passages"| node_openai
node_incdet -->|"analyses passages"| node_openai
node_workflow -->|"coordinates agents"| node_orchestrator
node_orchestrator -->|"delegates analysis"| node_dupagent
node_orchestrator -->|"delegates analysis"| node_confagent
node_orchestrator -->|"delegates analysis"| node_incagent
node_orchestrator -->|"uses instructions"| node_prompts
node_confagent -->|"uses instructions"| node_prompts
node_dupagent -->|"uses instructions"| node_prompts
node_incagent -->|"uses instructions"| node_prompts
node_confagent -->|"searches knowledge"| node_store
node_confagent -->|"checks conflicts"| node_confdet
node_workflow -->|"returns findings"| node_results
node_workflow -->|"summarizes findings"| node_summaries
node_dupdet -->|"produces findings"| node_results
node_confdet -->|"produces findings"| node_results
node_incdet -->|"produces findings"| node_results

click node_cli "https://github.com/dfe-digital/document-analysis-agentic-poc/blob/main/src/main.py"
click node_workflow "https://github.com/dfe-digital/document-analysis-agentic-poc/blob/main/src/agents/conflict_detector_agent.py"
click node_loader "https://github.com/dfe-digital/document-analysis-agentic-poc/blob/main/src/knowledge/document_loader.py"
click node_store "https://github.com/dfe-digital/document-analysis-agentic-poc/blob/main/src/knowledge/vector_store.py"
click node_dupdet "https://github.com/dfe-digital/document-analysis-agentic-poc/blob/main/src/analysis/duplication_detector.py"
click node_confdet "https://github.com/dfe-digital/document-analysis-agentic-poc/blob/main/src/analysis/conflict_detector.py"
click node_incdet "https://github.com/dfe-digital/document-analysis-agentic-poc/blob/main/src/analysis/inconsistency_detector.py"
click node_orchestrator "https://github.com/dfe-digital/document-analysis-agentic-poc/blob/main/src/agents/specialized/orchestrator_agent.py"
click node_dupagent "https://github.com/dfe-digital/document-analysis-agentic-poc/blob/main/src/agents/specialized/duplication_agent.py"
click node_confagent "https://github.com/dfe-digital/document-analysis-agentic-poc/blob/main/src/agents/specialized/conflict_agent.py"
click node_incagent "https://github.com/dfe-digital/document-analysis-agentic-poc/blob/main/src/agents/specialized/inconsistency_agent.py"
click node_prompts "https://github.com/dfe-digital/document-analysis-agentic-poc/blob/main/src/agents/prompts.py"
click node_results "https://github.com/dfe-digital/document-analysis-agentic-poc/blob/main/src/agents/models.py"
click node_summaries "https://github.com/dfe-digital/document-analysis-agentic-poc/blob/main/src/agents/one_liners.py"

classDef toneNeutral fill:#f8fafc,stroke:#334155,stroke-width:1.5px,color:#0f172a
classDef toneBlue fill:#dbeafe,stroke:#2563eb,stroke-width:1.5px,color:#172554
classDef toneAmber fill:#fef3c7,stroke:#d97706,stroke-width:1.5px,color:#78350f
classDef toneMint fill:#dcfce7,stroke:#16a34a,stroke-width:1.5px,color:#14532d
classDef toneRose fill:#ffe4e6,stroke:#e11d48,stroke-width:1.5px,color:#881337
classDef toneIndigo fill:#e0e7ff,stroke:#4f46e5,stroke-width:1.5px,color:#312e81
classDef toneTeal fill:#ccfbf1,stroke:#0f766e,stroke-width:1.5px,color:#134e4a
class node_cli,node_workflow,node_user toneBlue
class node_loader,node_store toneAmber
class node_dupdet,node_confdet,node_incdet,node_results,node_summaries toneMint
class node_orchestrator,node_dupagent,node_confagent,node_incagent,node_prompts toneRose
class node_openai toneIndigo
```

## Components

### User Entry

- [CLI Commands](https://github.com/dfe-digital/document-analysis-agentic-poc/blob/main/src/main.py): `src/main.py`
- [Analysis Facade](https://github.com/dfe-digital/document-analysis-agentic-poc/blob/main/src/agents/conflict_detector_agent.py): `src/agents/conflict_detector_agent.py`

### Knowledge Base

- [Document Loader](https://github.com/dfe-digital/document-analysis-agentic-poc/blob/main/src/knowledge/document_loader.py): `src/knowledge/document_loader.py`
- [Vector Store](https://github.com/dfe-digital/document-analysis-agentic-poc/blob/main/src/knowledge/vector_store.py): `src/knowledge/vector_store.py`

### Analysis

- [Duplication Detector](https://github.com/dfe-digital/document-analysis-agentic-poc/blob/main/src/analysis/duplication_detector.py): `src/analysis/duplication_detector.py`
- [Conflict Detector](https://github.com/dfe-digital/document-analysis-agentic-poc/blob/main/src/analysis/conflict_detector.py): `src/analysis/conflict_detector.py`
- [Inconsistency Detector](https://github.com/dfe-digital/document-analysis-agentic-poc/blob/main/src/analysis/inconsistency_detector.py): `src/analysis/inconsistency_detector.py`
- [Analysis Models](https://github.com/dfe-digital/document-analysis-agentic-poc/blob/main/src/agents/models.py): `src/agents/models.py`
- [Finding Summaries](https://github.com/dfe-digital/document-analysis-agentic-poc/blob/main/src/agents/one_liners.py): `src/agents/one_liners.py`

### Agent Workflow

- [Orchestrator Agent](https://github.com/dfe-digital/document-analysis-agentic-poc/blob/main/src/agents/specialized/orchestrator_agent.py): `src/agents/specialized/orchestrator_agent.py`
- [Duplication Agent](https://github.com/dfe-digital/document-analysis-agentic-poc/blob/main/src/agents/specialized/duplication_agent.py): `src/agents/specialized/duplication_agent.py`
- [Conflict Agent](https://github.com/dfe-digital/document-analysis-agentic-poc/blob/main/src/agents/specialized/conflict_agent.py): `src/agents/specialized/conflict_agent.py`
- [Inconsistency Agent](https://github.com/dfe-digital/document-analysis-agentic-poc/blob/main/src/agents/specialized/inconsistency_agent.py): `src/agents/specialized/inconsistency_agent.py`
- [Agent Prompts](https://github.com/dfe-digital/document-analysis-agentic-poc/blob/main/src/agents/prompts.py): `src/agents/prompts.py`

### Other components

- **Terminal User**
- **OpenAI Models**

## Connections

- Terminal User → CLI Commands: runs commands (evidence: `README.md`)
- CLI Commands → Analysis Facade: invokes analysis (evidence: `src/main.py`)
- CLI Commands → Analysis Facade: adds and searches (evidence: `README.md`)
- Analysis Facade → Document Loader: loads documents (evidence: `src/agents/conflict_detector_agent.py`)
- Analysis Facade → Vector Store: stores and retrieves (evidence: `src/agents/conflict_detector_agent.py`)
- Document Loader → Vector Store: provides chunks (evidence: `src/knowledge/vector_store.py`)
- Analysis Facade → Duplication Detector: runs detection (evidence: `src/agents/conflict_detector_agent.py`)
- Analysis Facade → Conflict Detector: runs detection (evidence: `src/agents/conflict_detector_agent.py`)
- Analysis Facade → Inconsistency Detector: runs detection (evidence: `src/agents/conflict_detector_agent.py`)
- Duplication Detector → Vector Store: searches similar chunks (evidence: `src/analysis/duplication_detector.py`)
- Conflict Detector → Vector Store: retrieves candidates (evidence: `src/analysis/conflict_detector.py`)
- Inconsistency Detector → Vector Store: uses semantic search (evidence: `src/analysis/inconsistency_detector.py`)
- Conflict Detector → OpenAI Models: analyses passages (evidence: `src/analysis/conflict_detector.py`)
- Inconsistency Detector → OpenAI Models: analyses passages (evidence: `src/analysis/inconsistency_detector.py`)
- Analysis Facade → Orchestrator Agent: coordinates agents (evidence: `README.md`)
- Orchestrator Agent → Duplication Agent: delegates analysis (evidence: `README.md`)
- Orchestrator Agent → Conflict Agent: delegates analysis (evidence: `README.md`)
- Orchestrator Agent → Inconsistency Agent: delegates analysis (evidence: `README.md`)
- Orchestrator Agent → Agent Prompts: uses instructions (evidence: `src/agents/specialized/orchestrator_agent.py`)
- Conflict Agent → Agent Prompts: uses instructions (evidence: `src/agents/specialized/conflict_agent.py`)
- Duplication Agent → Agent Prompts: uses instructions (evidence: `src/agents/specialized/duplication_agent.py`)
- Inconsistency Agent → Agent Prompts: uses instructions (evidence: `src/agents/specialized/inconsistency_agent.py`)
- Conflict Agent → Vector Store: searches knowledge (evidence: `src/agents/specialized/conflict_agent.py`)
- Conflict Agent → Conflict Detector: checks conflicts (evidence: `src/agents/specialized/conflict_agent.py`)
- Analysis Facade → Analysis Models: returns findings (evidence: `src/agents/conflict_detector_agent.py`)
- Analysis Facade → Finding Summaries: summarizes findings (evidence: `src/agents/conflict_detector_agent.py`)
- Duplication Detector → Analysis Models: produces findings (evidence: `src/agents/models.py`)
- Conflict Detector → Analysis Models: produces findings (evidence: `src/agents/models.py`)
- Inconsistency Detector → Analysis Models: produces findings (evidence: `src/agents/models.py`)
