Wiki source code of AI Knowledge Extraction Layer (AKEL)
Version 2.1 by Robert Schaub on 2025/12/11 21:35
Show last authors
| author | version | line-number | content |
|---|---|---|---|
| 1 | = AKEL — AI Knowledge Extraction Layer = | ||
| 2 | |||
| 3 | AKEL is FactHarbor’s automated intelligence subsystem. | ||
| 4 | Its purpose is to reduce human workload, enhance consistency, and enable scalable knowledge processing — **without ever replacing human judgment**. | ||
| 5 | |||
| 6 | All AKEL outputs are marked with **AuthorType = AI** and require human approval before publication. | ||
| 7 | |||
| 8 | AKEL operates in two modes: | ||
| 9 | |||
| 10 | * **Single-node mode** (POC & Beta 0) | ||
| 11 | * **Federated multi-node mode** (Release 1.0+) | ||
| 12 | |||
| 13 | Human reviewers, experts, and moderators always retain final authority. | ||
| 14 | |||
| 15 | ---- | ||
| 16 | |||
| 17 | == Purpose and Role == | ||
| 18 | |||
| 19 | AKEL transforms unstructured inputs into structured, review-ready drafts. | ||
| 20 | |||
| 21 | Core responsibilities: | ||
| 22 | |||
| 23 | * Claim extraction from arbitrary text | ||
| 24 | * Claim classification (domain, type, evaluability, safety) | ||
| 25 | * Scenario generation (definitions, boundaries, assumptions, methodology) | ||
| 26 | * Evidence summarization and metadata extraction | ||
| 27 | * Contradiction detection | ||
| 28 | * Re-evaluation proposal generation | ||
| 29 | * Cross-node embedding exchange (Release 1.0+) | ||
| 30 | |||
| 31 | ---- | ||
| 32 | |||
| 33 | == Components == | ||
| 34 | |||
| 35 | * **AKEL Orchestrator** – central coordinator | ||
| 36 | * **Claim Extractor** | ||
| 37 | * **Claim Classifier** | ||
| 38 | * **Scenario Generator** | ||
| 39 | * **Evidence Summarizer** | ||
| 40 | * **Contradiction Detector** | ||
| 41 | * **Embedding Handler** (Release 1.0+) | ||
| 42 | * **Federation Sync Adapter** (Release 1.0+) | ||
| 43 | |||
| 44 | ---- | ||
| 45 | |||
| 46 | == Inputs and Outputs == | ||
| 47 | |||
| 48 | === Inputs === | ||
| 49 | * User-submitted claims or evidence | ||
| 50 | * Uploaded documents | ||
| 51 | * URLs or citations | ||
| 52 | * External LLM API (optional) | ||
| 53 | * Embeddings (from local or federated peers) | ||
| 54 | |||
| 55 | === Outputs (all require human approval) === | ||
| 56 | * ClaimVersion (draft) | ||
| 57 | * ScenarioVersion (draft) | ||
| 58 | * EvidenceVersion (summary + metadata draft) | ||
| 59 | * VerdictVersion (draft; internal only) | ||
| 60 | * Contradiction alerts | ||
| 61 | * Re-evaluation proposals | ||
| 62 | * Updated embeddings | ||
| 63 | |||
| 64 | ---- | ||
| 65 | |||
| 66 | == Architecture Overview == | ||
| 67 | |||
| 68 | {{include reference="FactHarbor.Specification.Diagrams.AKEL Architecture.WebHome"/}} | ||
| 69 | |||
| 70 | ---- | ||
| 71 | |||
| 72 | == AKEL and Federation == | ||
| 73 | |||
| 74 | In Release 1.0+, AKEL participates in cross-node knowledge alignment: | ||
| 75 | |||
| 76 | * Shares embeddings | ||
| 77 | * Exchanges canonicalized claim forms | ||
| 78 | * Exchanges scenario templates | ||
| 79 | * Sends + receives contradiction alerts | ||
| 80 | * Never shares model weights | ||
| 81 | * Never overrides local governance | ||
| 82 | |||
| 83 | Nodes may choose trust levels for AKEL-related data: | ||
| 84 | |||
| 85 | * Trusted nodes: auto-merge embeddings + templates | ||
| 86 | * Neutral nodes: require reviewer approval | ||
| 87 | * Untrusted nodes: fully manual import | ||
| 88 | |||
| 89 | ---- | ||
| 90 | |||
| 91 | == Human Approval Workflow == | ||
| 92 | |||
| 93 | 1. AKEL generates draft outputs (AuthorType = AI) | ||
| 94 | 2. Reviewers inspect and approve/moderate the drafts | ||
| 95 | 3. Experts validate high-risk or domain-specific outputs | ||
| 96 | 4. Moderators finalize publication | ||
| 97 | 5. Version numbers increment, history preserved | ||
| 98 | |||
| 99 | No AKEL output is ever published automatically. | ||
| 100 | |||
| 101 | ---- |