Document Analysis scaffold

A complete program, not a fragment: it compiles as written. Copy it, rename the placeholder identifiers to your domain, and run axon check — the compiler will tell you what your renaming broke. An MCP client can also generate it through the axon.compose tool by naming the document_analysis domain.

// AXON Document Analysis scaffold — bulk document ingestion +
// entity / clause / topic extraction for downstream Q&A.
//
// Differs from `chat_research.axon` (corpus-grounded chat) and
// `knowledge_extraction.axon` (entity-graph focus): this scaffold
// targets the GENERAL bulk-document-processing pattern (invoices,
// resumes, reports, technical docs).

// ── Types ─────────────────────────────────────────────────────────

type SourceDocument {
    document_id:    Text
    mime_type:      Text
    content:        Text
    received_at:    Text
}

type DocumentAnalysis {
    document_id:    Text
    summary:        Text
    extracted_data: Text
    confidence:     Number
    classification: Text
}

type AnalyzeDocumentRequest { doc: SourceDocument }

// ── Identity + grounding ──────────────────────────────────────────

persona DocumentAnalyst {
    domain: ["document-processing", "information-extraction", "text-analytics"]
    tone: precise
    confidence_threshold: 0.8
    cite_sources: true
}

context DocumentReview {
    memory: session
    language: "en"
    depth: deep
    max_tokens: 4096
    temperature: 0.15
}

anchor NoUnverifiedExtraction {
    require: source_citation
    confidence_floor: 0.75
    unknown_response: "Extraction confidence below threshold — routing to manual review."
    on_violation: log
}

// ── Shield ────────────────────────────────────────────────────────

shield DocumentShield {
    scan:       [prompt_injection, pii_leak, data_exfil]
    on_breach:  quarantine
    severity:   high
    compliance: [SOC2]
}

// ── Flow ──────────────────────────────────────────────────────────

flow AnalyzeDocument(doc: SourceDocument) -> FlowEnvelope<DocumentAnalysis> {
    step Classify {
        given: doc
        ask: "Classify the document type (invoice / resume / report / other); emit a confidence score."
        output: FlowEnvelope<DocumentAnalysis>
    }
    step Extract {
        given: Classify.output
        ask: "Extract the typed structured data for this document class; cite the source span for each field."
        output: FlowEnvelope<DocumentAnalysis>
    }
    return Extract.output
}

// ── HTTP boundary ─────────────────────────────────────────────────

axonendpoint DocumentAnalysisAPI {
    method:      post
    path:        "/v1/documents/analyze"
    body:        AnalyzeDocumentRequest
    execute:     AnalyzeDocument
    output:      FlowEnvelope<DocumentAnalysis>
    shield:      DocumentShield
    backend:     auto
    compliance:  [SOC2]
    retries:     1
    timeout:     45s
    requires:    ["documents.analyze"]
}