Skip to Content
Last modified by Robert Schaub on 2026/02/08 08:12
From version 5.1
edited by Robert Schaub
on 2026/01/02 10:04
Change comment: Rollback to version 3.1
To version 2.1
edited by Robert Schaub
on 2026/01/02 10:01
Change comment: There is no comment for this version

Summary

Details

Page properties
Content
... ... @@ -5,6 +5,8 @@
5 5  **Analysis Date:** January 2026
6 6  **Document Purpose:** Technical diagrams, gap analysis, and optimization recommendations
7 7  
8 +-----
9 +
8 8  ----
9 9  
10 10  == 1. AKEL Flow Diagram (with LLM and WebSearch Interactions) ==
... ... @@ -91,7 +91,7 @@
91 91   class UNDERSTAND,DECIDE,FETCHSRC,EXTRACT,VERDICT,REPORT step
92 92  {{/mermaid}}
93 93  
94 -----
96 +-----
95 95  
96 96  
97 97  == 2. ERD Data Model (Current POC1 Implementation) ==
... ... @@ -186,7 +186,7 @@
186 186   }
187 187  {{/mermaid}}
188 188  
189 -----
191 +-----
190 190  
191 191  
192 192  == 3. Overall Architecture with Interactions ==
... ... @@ -285,7 +285,7 @@
285 285   class ANALYZE_API,JOBS_API,JOB_API,EVENTS_API,RUN_JOB api
286 286  {{/mermaid}}
287 287  
288 -----
290 +-----
289 289  
290 290  
291 291  == 4. Specification vs Implementation Gap Analysis ==
... ... @@ -324,7 +324,7 @@
324 324  
325 325  
326 326  | Spec Requirement | POC1 Status | Gap Description |
327 -| |-|-|
329 +||-|-|
328 328  | **Three-Layer Architecture** | ✅ Implemented | Interface (Next.js) → Processing (AKEL) → Data (SQLite) |
329 329  | **LLM Abstraction Layer** | ✅ Implemented | AI SDK supports multiple providers with failover |
330 330  | **PostgreSQL Primary DB** | ⚠️ Different | Using SQLite for simplicity (acceptable for POC) |
... ... @@ -337,13 +337,13 @@
337 337  
338 338  
339 339  | Spec Feature | POC1 Status | Gap Description |
340 -| |-|-|
342 +||-|-|
341 341  | **Risk Tier Publication Rules** | ❌ Missing | All results published immediately regardless of tier |
342 342  | **Human Review Queue** | ❌ Missing | No review workflow |
343 343  | **AI-Generated Labeling** | ⚠️ Partial | Results show "AI analysis" but no formal labeling system |
344 344  | **Audit Rate Sampling** | ❌ Missing | No sampling audits |
345 345  
346 -----
348 +-----
347 347  
348 348  
349 349  == 5. Optimization Recommendations ==
... ... @@ -361,7 +361,7 @@
361 361  {{/mermaid}}
362 362  
363 363  | Optimization | Estimated Savings | Implementation Effort |
364 -| |-| |
366 +||-----||
365 365  | **Cache claim understanding** | 30-50% on repeated claims | Medium |
366 366  | **Use Haiku for fact extraction** | 40% on Step 2 costs | Low (config change) |
367 367  | **Batch fact extraction** | 20% fewer API calls | Medium |
... ... @@ -397,7 +397,7 @@
397 397  {{/mermaid}}
398 398  
399 399  | Optimization | Time Savings | Notes |
400 -| | |-|
402 +|||-----|
401 401  | **Parallel source fetching** | Already implemented | Currently fetches 3 sources in parallel |
402 402  | **Streaming LLM responses** | 20-30% perceived | User sees progress faster |
403 403  | **Search query batching** | 10-15% | Send multiple queries to search API |
... ... @@ -422,7 +422,7 @@
422 422   - Cache search results (1h TTL)
423 423   - Reduces external API calls
424 424  
425 -----
427 +-----
426 426  
427 427  
428 428  == 6. Separated Verdict Architecture Proposal ==
... ... @@ -508,7 +508,7 @@
508 508  
509 509  
510 510  | Benefit | Impact | Rationale |
511 -|-| |-|
513 +|-| |-----|
512 512  | **Cost Reduction** | 40-70% for repeated claims | Many articles share common claims (e.g., "COVID vaccines are safe") |
513 513  | **Faster Analysis** | 50%+ for cached claims | Skip research + LLM calls for known claims |
514 514  | **Consistency** | High | Same claim always gets same verdict (until cache expires) |
... ... @@ -552,7 +552,7 @@
552 552   - Phase 2: Semantic similarity caching (embedding-based)
553 553   - Phase 3: Federated claim sharing across instances
554 554  
555 -----
557 +-----
556 556  
557 557  
558 558  == 7. Summary ==