AKEL Pipeline Flow

High-Level Flow

flowchart TB
    subgraph Input["📥 Input Layer"]
        URL[URL Input]
        TEXT[Text Input]
    end

    subgraph Retrieval["🔍 Content Retrieval"]
        FETCH[extractTextFromUrl]
        PDF[PDF Parser
pdf2json]
        HTML[HTML Parser
cheerio]
    end

    subgraph AKEL["🧠 AKEL Pipeline"]
        direction TB

        subgraph Step1["Step 1: Understand"]
            UNDERSTAND[understandClaim
━━━━━━━━━━━━━
• Detect input type
• Extract claims
• Identify dependencies
• Assign risk tiers]
            LLM1[("🤖 LLM Call #1
Claude/GPT/Gemini")]
        end

        subgraph Step2["Step 2: Research (Iterative)"]
            DECIDE[decideNextResearch
━━━━━━━━━━━━━
• Generate queries
• Focus areas]

            SEARCH[("🌐 Web Search
Google CSE / SerpAPI")]

            FETCHSRC[fetchSourceContent
━━━━━━━━━━━━━
• Parallel fetching
• Timeout handling]

            EXTRACT[extractEvidence
━━━━━━━━━━━━━
• Parse sources
• Extract evidence]
            LLM2[("🤖 LLM Call #2-N
Per source")]
        end

        subgraph Step3["Step 3: Verdict Generation"]
            VERDICT[generateVerdicts
━━━━━━━━━━━━━
• Claim verdicts
• Article verdict
• Dependency propagation]
            LLM3[("🤖 LLM Call #N+1
Final synthesis")]
        end

        subgraph Step4["Step 4: Summary"]
            SUMMARY[generateTwoPanelSummary
━━━━━━━━━━━━━
• Format results
• Build two-panel summary]
        end

        subgraph Step5["Step 5: Report"]
            REPORT[generateReport
━━━━━━━━━━━━━
• Generate markdown]
        end
    end

    subgraph Output["📤 Output"]
        RESULT[AnalysisResult JSON]
        MARKDOWN[Report Markdown]
    end

    %% Flow connections
    URL --> FETCH
    TEXT --> UNDERSTAND
    FETCH --> PDF
    FETCH --> HTML
    PDF --> UNDERSTAND
    HTML --> UNDERSTAND

    UNDERSTAND --> LLM1
    LLM1 --> DECIDE

    DECIDE --> SEARCH
    SEARCH --> FETCHSRC
    FETCHSRC --> EXTRACT
    EXTRACT --> LLM2
    LLM2 --> DECIDE

    DECIDE -->|"Research Complete"| VERDICT
    VERDICT --> LLM3
    LLM3 --> SUMMARY
    SUMMARY --> REPORT

    REPORT --> RESULT
    REPORT --> MARKDOWN

Pipeline Steps Detail

Step 1: Understand (understandClaim)

• Detects input type: question | statement | article
• Extracts claims with dependencies
• Assigns risk tiers (A/B/C)
• Detects context(s) and temporal boundaries
• Discovers KeyFactors (optional decomposition questions)
• Applies Gate 1: Claim Validation

Step 2: Research (decideNextResearch + extractEvidence)

• Iterative research cycle (typically 2-3 rounds)
• Generates search queries targeting gaps
• Fetches and parses sources (HTML, PDF)
• Extracts evidence from each source
• Continues until research is complete or max rounds reached

Step 3: Verdict Generation (generateVerdicts)

• Generates verdicts for each claim
• Aggregates claim verdicts into KeyFactor verdicts
• Aggregates KeyFactor verdicts into context answers
• Generates overall article verdict
• Applies Gate 4: Verdict Confidence Assessment

Step 4: Summary (generateTwoPanelSummary)

• Builds two-panel summary (Overview + Key Findings)
• Formats verdict data for display

Step 5: Report (generateReport)

• Generates markdown report
• Includes all sections: Summary, Claims, Sources, Verdict

Navigation: Architecture Overview | Prev: Architecture Summary | Next: Data Models