Skip to Content

Changes for page AKEL Architecture Mermaid

Last modified by Robert Schaub on 2025/12/24 21:53
From version 2.1
edited by Robert Schaub
on 2025/12/18 22:28
Change comment: Imported from XAR
To version 1.1
edited by Robert Schaub
on 2025/12/18 12:03
Change comment: Imported from XAR

Summary

Details

Page properties
Content
... ... @@ -1,26 +1,29 @@
1 1  {{mermaid}}
2 2  graph TB
3 - User[User Submits Content<br/>Text/URL/Single Claim]
4 - Extract[Claim Extraction<br/>LLM identifies distinct claims]
5 - AKEL[AKEL Core Processing<br/>Per Claim]
6 - Evidence[Evidence Gathering]
7 - Scenario[Scenario Generation]
8 - Verdict[Verdict Generation]
9 - Storage[(Storage Layer<br/>PostgreSQL + S3)]
10 - Queue[Processing Queue<br/>Parallel Claims]
11 -
12 - User --> Extract
13 - Extract -->|Multiple Claims| Queue
14 - Extract -->|Single Claim| AKEL
15 - Queue -->|Process Each| AKEL
16 -
17 - AKEL --> Evidence
18 - AKEL --> Scenario
19 - Evidence --> Verdict
20 - Scenario --> Verdict
21 - Verdict --> Storage
22 -
23 - style Extract fill:#e1f5ff
24 - style Queue fill:#fff4e1
25 - style AKEL fill:#f0f0f0
3 + START[User Submits Claim] --> PARSE[Parse & Extract<br/>Components<br/>2-3 seconds]
4 + PARSE --> PARALLEL{Parallel<br/>Processing}
5 + PARALLEL -->|Branch 1| EVIDENCE[Evidence Gathering<br/>Web + DB search<br/>5-10 seconds]
6 + PARALLEL -->|Branch 2| SOURCES[Source Checking<br/>Track record lookup<br/>2-3 seconds]
7 + PARALLEL -->|Branch 3| SCENARIOS[Extract Scenarios<br/>From evidence context<br/>3-5 seconds]
8 + EVIDENCE --> VERDICT[Synthesize Verdict<br/>Compile evidence per scenario<br/>2-3 seconds]
9 + SOURCES --> VERDICT
10 + SCENARIOS --> VERDICT
11 + VERDICT --> RISK[Calculate Risk<br/>Assess harm & controversy<br/>1-2 seconds]
12 + RISK --> PUBLISH[Publish Immediately<br/>No approval gates]
13 + PUBLISH --> METRICS[Update Metrics<br/>Background]
14 + LLM[LLM APIs<br/>Claude/GPT/Local<br/>Orchestration Layer] -.->|Used by| PARSE
15 + LLM -.->|Used by| EVIDENCE
16 + LLM -.->|Used by| SCENARIOS
17 + LLM -.->|Used by| VERDICT
18 + LLM -.->|Used by| RISK
19 + style PARSE fill:#ffcccc
20 + style EVIDENCE fill:#ccffcc
21 + style SOURCES fill:#ccffcc
22 + style SCENARIOS fill:#ccffcc
23 + style VERDICT fill:#ffcccc
24 + style RISK fill:#ffcccc
25 + style PUBLISH fill:#ccccff
26 + style LLM fill:#ffffcc
26 26  {{/mermaid}}
28 +
29 +**AKEL Pipeline with Parallel Processing** - Parse claim (2-3s), then process evidence/sources/scenarios in parallel (max 5-10s), synthesize verdict (2-3s), calculate risk (1-2s). Total: 10-18 seconds (vs 15-30s sequential). Multiple LLM providers supported via orchestration layer for redundancy and cross-checking. No approval gates - immediate publication.