Wiki source code of POC Requirements

Last modified by Robert Schaub on 2025/12/22 14:38

Show last authors
1 = POC Requirements = **Status:** ✅ Approved for Development **Version:** 3.0 (Aligned with Main Requirements) **Goal:** Prove that AI can extract claims and determine verdicts automatically without human intervention {{info}}
2 **Core Philosophy:** POC validates the [[Main Requirements>>FactHarbor.Specification.Requirements.WebHome]] through simplified implementation. All POC features map to formal FR/NFR requirements.
3 {{/info}} == 1. POC Overview == === 1.1 What POC Tests === **Core Question:**
4 > Can AI automatically extract factual claims from articles and evaluate them with reasonable verdicts? **What we're proving:**
5 * AI can identify factual claims from text
6 * AI can evaluate those claims with structured evidence
7 * Quality gates can filter unreliable outputs
8 * The core workflow is technically feasible **What we're NOT proving:**
9 * Production-ready reliability (that's POC2)
10 * User-facing features (that's Beta 0)
11 * Full IFCN compliance (that's V1.0) === 1.2 Requirements Mapping === POC1 implements a **subset** of the full system requirements defined in [[Main Requirements>>FactHarbor.Specification.Requirements.WebHome]]. **Scope Summary:**
12 * **In Scope:** 8 requirements (7 FRs + 1 NFR)
13 * **Partial:** 3 NFRs (simplified versions)
14 * **Out of Scope:** 19 requirements (deferred to later phases) == 2. Requirements Scope Matrix == {{success}}
15 **Requirements Traceability:** This matrix shows which [[Main Requirements>>FactHarbor.Specification.Requirements.WebHome]] are implemented in POC1, providing full traceability between POC and system requirements.
16 {{/success}} |=Requirement|=POC1 Status|=Implementation Level|=Notes
17 |**CORE WORKFLOW**||||
18 |FR1: Claim Extraction|✅ **In Scope**|Full|AKEL extracts claims from text
19 |FR2: Claim Context|✅ **In Scope**|Basic|Context preserved with claim
20 |FR3: Multiple Scenarios|✅ **In Scope**|Full|AKEL generates interpretation scenarios
21 |FR4: Analysis Summary|✅ **In Scope**|Basic|Simple summary format
22 |FR5: Evidence Collection|✅ **In Scope**|Full|AKEL searches for evidence
23 |FR6: Evidence Evaluation|✅ **In Scope**|Full|AKEL evaluates source reliability
24 |FR7: Automated Verdicts|✅ **In Scope**|Full|AKEL computes verdicts with uncertainty
25 |**QUALITY & RELIABILITY**||||
26 |NFR11: Quality Assurance|✅ **In Scope**|**Lite**|**2 gates only** (Gate 1 & 4)
27 |NFR1: Performance|⚠️ **Partial**|Basic|Response time monitored, not optimized
28 |NFR2: Scalability|⚠️ **Partial**|Single-thread|No concurrent processing
29 |NFR3: Reliability|⚠️ **Partial**|Basic|Error handling, no retry logic
30 |**DEFERRED TO LATER**||||
31 |FR8-FR13|❌ Out of Scope|N/A|User accounts, corrections, publishing
32 |FR44-FR53|❌ Out of Scope|N/A|Advanced features (V1.0+)
33 |NFR4: Security|❌ Out of Scope|N/A|POC2
34 |NFR5: Maintainability|❌ Out of Scope|N/A|POC2
35 |NFR12: Security Controls|❌ Out of Scope|N/A|Beta 0
36 |NFR13: Monitoring|❌ Out of Scope|N/A|POC2 == 3. POC Simplifications == === 3.1 FR1: Claim Extraction (Full Implementation) === **Main Requirement:** AI extracts factual claims from input text **POC Implementation:**
37 * ✅ AKEL extracts claims using LLM
38 * ✅ Each claim includes original text reference
39 * ✅ Claims are identified as factual/non-factual
40 * ❌ No advanced claim parsing (added in POC2) **Acceptance Criteria:**
41 * Extracts 3-5 claims from typical article
42 * Identifies factual vs non-factual claims
43 * Quality Gate 1 validates extraction === 3.2 FR3: Multiple Scenarios (Full Implementation) === **Main Requirement:** Generate multiple interpretation scenarios for ambiguous claims **POC Implementation:**
44 * ✅ AKEL generates 2-3 scenarios per claim
45 * ✅ Scenarios capture different interpretations
46 * ✅ Each scenario is evaluated separately
47 * ✅ Verdict considers all scenarios **Acceptance Criteria:**
48 * Generates 2+ scenarios for ambiguous claims
49 * Scenarios are meaningfully different
50 * All scenarios are evaluated === 3.3 FR4: Analysis Summary (Basic Implementation) === **Main Requirement:** Provide user-friendly summary of analysis **POC Implementation:**
51 * ✅ Simple text summary generated
52 * ❌ No rich formatting (added in Beta 0)
53 * ❌ No visual elements (added in Beta 0)
54 * ❌ No interactive features (added in Beta 0) **POC Format:**
55 ```
56 Claim: [extracted claim]
57 Scenarios: [list of scenarios]
58 Evidence: [supporting/opposing evidence]
59 Verdict: [probability with uncertainty]
60 ``` === 3.4 FR5-FR6: Evidence Collection & Evaluation (Full Implementation) === **Main Requirements:**
61 * FR5: Collect supporting and opposing evidence
62 * FR6: Evaluate evidence source reliability **POC Implementation:**
63 * ✅ AKEL searches for evidence (web/knowledge base)
64 * ✅ **Mandatory contradiction search** (finds opposing evidence)
65 * ✅ Source reliability scoring
66 * ❌ No evidence deduplication (added in POC2)
67 * ❌ No advanced source verification (added in POC2) **Acceptance Criteria:**
68 * Finds 2+ supporting evidence items
69 * Finds 1+ opposing evidence (if exists)
70 * Sources scored for reliability === 3.5 FR7: Automated Verdicts (Full Implementation) === **Main Requirement:** AI computes verdicts with uncertainty quantification **POC Implementation:**
71 * ✅ Probabilistic verdicts (0-100% confidence)
72 * ✅ Uncertainty explicitly stated
73 * ✅ Reasoning chain provided
74 * ✅ Quality Gate 4 validates verdict confidence **POC Output:**
75 ```
76 Verdict: 70% likely true
77 Uncertainty: ±15% (moderate confidence)
78 Reasoning: Based on 3 high-quality sources...
79 Confidence Level: MEDIUM
80 ``` **Acceptance Criteria:**
81 * Verdicts include probability (0-100%)
82 * Uncertainty explicitly quantified
83 * Reasoning chain explains verdict === 3.6 NFR11: Quality Assurance Framework (LITE VERSION) === **Main Requirement:** Complete quality assurance with 7 quality gates **POC Implementation:** **2 gates only** **Quality Gate 1: Claim Validation**
84 * ✅ Validates claim is factual and verifiable
85 * ✅ Blocks non-factual claims (opinion/prediction/ambiguous)
86 * ✅ Provides clear rejection reason **Quality Gate 4: Verdict Confidence Assessment**
87 * ✅ Validates ≥2 sources found
88 * ✅ Validates quality score ≥0.6
89 * ✅ Blocks low-confidence verdicts
90 * ✅ Provides clear rejection reason **Out of Scope (POC2+):**
91 * ❌ Gate 2: Evidence Relevance
92 * ❌ Gate 3: Scenario Coherence
93 * ❌ Gate 5: Source Diversity
94 * ❌ Gate 6: Reasoning Validity
95 * ❌ Gate 7: Output Completeness **Rationale:** Prove gate concept works. Add remaining gates in POC2 after validating approach. === 3.7 NFR1-3: Performance, Scalability, Reliability (Basic) === **Main Requirements:**
96 * NFR1: Response time < 30 seconds
97 * NFR2: Handle 1000+ concurrent users
98 * NFR3: 99.9% uptime **POC Implementation:**
99 * ⚠️ **Response time monitored** (not optimized)
100 * ⚠️ **Single-threaded processing** (no concurrency)
101 * ⚠️ **Basic error handling** (no advanced retry logic) **Rationale:** POC proves functionality. Performance optimization happens in POC2. **POC Acceptance:**
102 * Analysis completes (no timeout requirement)
103 * Errors don't crash system
104 * Basic logging in place == 4. What's NOT in POC Scope == === 4.1 User-Facing Features (Beta 0+) === {{warning}}
105 **Deferred to Beta 0:**
106 {{/warning}} **Out of Scope:**
107 * ❌ User accounts and authentication (FR8)
108 * ❌ User corrections system (FR9, FR45-46)
109 * ❌ Public publishing interface (FR10)
110 * ❌ Social sharing (FR11)
111 * ❌ Email notifications (FR12)
112 * ❌ API access (FR13) **Rationale:** POC validates AI capabilities. User features added in Beta 0. === 4.2 Advanced Features (V1.0+) === **Out of Scope:**
113 * ❌ IFCN compliance (FR47)
114 * ❌ ClaimReview schema (FR48)
115 * ❌ Archive.org integration (FR49)
116 * ❌ OSINT toolkit (FR50)
117 * ❌ Video verification (FR51)
118 * ❌ Deepfake detection (FR52)
119 * ❌ Cross-org sharing (FR53) **Rationale:** Advanced features require proven platform. Added post-V1.0. === 4.3 Production Requirements (POC2, Beta 0) === **Out of Scope:**
120 * ❌ Security controls (NFR4, NFR12)
121 * ❌ Code maintainability (NFR5)
122 * ❌ System monitoring (NFR13)
123 * ❌ Evidence deduplication
124 * ❌ Advanced source verification
125 * ❌ Full 7-gate quality framework **Rationale:** POC proves concept. Production hardening happens in POC2 and Beta 0. == 5. POC Output Specification == === 5.1 Required Output Elements === For each analyzed claim, POC must produce: **1. Claim**
126 * Original text
127 * Classification (factual/non-factual/ambiguous)
128 * If non-factual: Clear reason why **2. Scenarios** (if factual)
129 * 2-3 interpretation scenarios
130 * Each scenario clearly described **3. Evidence** (if factual)
131 * Supporting evidence (2+ items)
132 * Opposing evidence (if exists)
133 * Source URLs and reliability scores **4. Verdict** (if factual)
134 * Probability (0-100%)
135 * Uncertainty quantification
136 * Confidence level (LOW/MEDIUM/HIGH)
137 * Reasoning chain **5. Quality Status**
138 * Which gates passed/failed
139 * If failed: Clear explanation why === 5.2 Example POC Output === {{code language="json"}}
140 { "claim": { "text": "Switzerland has the highest life expectancy in Europe", "type": "factual", "gate1_status": "PASS" }, "scenarios": [ "Switzerland's overall life expectancy is highest", "Switzerland ranks highest for specific age groups" ], "evidence": { "supporting": [ { "source": "WHO Report 2023", "reliability": 0.95, "excerpt": "Switzerland: 83.4 years average..." } ], "opposing": [ { "source": "Eurostat 2024", "reliability": 0.90, "excerpt": "Spain leads at 83.5 years..." } ] }, "verdict": { "probability": 0.65, "uncertainty": 0.15, "confidence": "MEDIUM", "reasoning": "WHO and Eurostat show similar but conflicting data...", "gate4_status": "PASS" }
141 }
142 {{/code}} == 6. Success Criteria == {{success}}
143 **POC Success Definition:** POC validates that AI can extract claims, find balanced evidence, and compute reasonable verdicts with quality gates improving output quality.
144 {{/success}} === 6.1 Functional Success === POC is successful if: ✅ **FR1-FR7 Requirements Met:**
145 1. Extracts 3-5 factual claims from test articles
146 2. Generates 2-3 scenarios per ambiguous claim
147 3. Finds supporting AND opposing evidence
148 4. Computes probabilistic verdicts with uncertainty
149 5. Provides clear reasoning chains ✅ **Quality Gates Work:**
150 1. Gate 1 blocks non-factual claims (100% block rate)
151 2. Gate 4 blocks low-quality verdicts (blocks if <2 sources or quality <0.6)
152 3. Clear rejection reasons provided ✅ **NFR11 Met:**
153 1. Quality gates reduce hallucination rate
154 2. Blocked outputs have clear explanations
155 3. Quality metrics are logged === 6.2 Quality Thresholds === **Minimum Acceptable:**
156 * ≥70% of test claims correctly classified (factual/non-factual)
157 * ≥60% of verdicts are reasonable (human evaluation)
158 * Gate 1 blocks 100% of non-factual claims
159 * Gate 4 blocks verdicts with <2 sources **Target:**
160 * ≥80% claims correctly classified
161 * ≥75% verdicts are reasonable
162 * <10% false positives (blocking good claims) === 6.3 POC Decision Gate === **After POC1, we decide:** **✅ PROCEED to POC2** if:
163 * Success criteria met
164 * Quality gates demonstrably improve output
165 * Core workflow is technically sound
166 * Clear path to production quality **⚠️ ITERATE POC1** if:
167 * Success criteria partially met
168 * Gates work but need tuning
169 * Core issues identified but fixable **❌ PIVOT APPROACH** if:
170 * Success criteria not met
171 * Fundamental AI limitations discovered
172 * Quality gates insufficient
173 * Alternative approach needed == 7. Test Cases == === 7.1 Happy Path === **Test 1: Simple Factual Claim**
174 * Input: "Paris is the capital of France"
175 * Expected: Factual, 1 scenario, verdict ~95% true **Test 2: Ambiguous Claim**
176 * Input: "Switzerland has the highest income in Europe"
177 * Expected: Factual, 2-3 scenarios, verdict with uncertainty **Test 3: Statistical Claim**
178 * Input: "10% of people have condition X"
179 * Expected: Factual, evidence with numbers, probabilistic verdict === 7.2 Edge Cases === **Test 4: Opinion**
180 * Input: "Paris is the best city"
181 * Expected: Non-factual (opinion), blocked by Gate 1 **Test 5: Prediction**
182 * Input: "Bitcoin will reach $100,000 next year"
183 * Expected: Non-factual (prediction), blocked by Gate 1 **Test 6: Insufficient Evidence**
184 * Input: Obscure factual claim with no sources
185 * Expected: Blocked by Gate 4 (<2 sources) === 7.3 Quality Gate Tests === **Test 7: Gate 1 Effectiveness**
186 * Input: Mix of 10 factual + 10 non-factual claims
187 * Expected: Gate 1 blocks all 10 non-factual (100% precision) **Test 8: Gate 4 Effectiveness**
188 * Input: Claims with varying evidence availability
189 * Expected: Gate 4 blocks low-confidence verdicts == 8. Technical Architecture (POC) == === 8.1 Simplified Architecture === **POC Tech Stack:**
190 * **Frontend:** Simple web interface (Next.js + TypeScript)
191 * **Backend:** Single API endpoint
192 * **AI:** Claude API (Sonnet 4.5)
193 * **Database:** Local JSON files (no database)
194 * **Deployment:** Single server **Architecture Diagram:** See [[POC1 Specification>>FactHarbor.Specification.POC.Specification]] === 8.2 AKEL Implementation === **POC AKEL:**
195 * Single-threaded processing
196 * Synchronous API calls
197 * No caching
198 * Basic error handling
199 * Console logging **Full AKEL (POC2+):**
200 * Multi-threaded processing
201 * Async API calls
202 * Evidence caching
203 * Advanced error handling with retry
204 * Structured logging + monitoring == 9. POC Philosophy == {{info}}
205 **Important:** POC validates concept, not production readiness. Focus is on proving AI can do the job, with production quality coming in later phases.
206 {{/info}} === 9.1 Core Principles === **1. Prove Concept, Not Production**
207 * POC validates AI can do the job
208 * Production quality comes in POC2 and Beta 0
209 * Focus on "does it work?" not "is it perfect?" **2. Implement Subset of Requirements**
210 * POC covers FR1-7, NFR11 (lite)
211 * All other requirements deferred
212 * Clear mapping to [[Main Requirements>>FactHarbor.Specification.Requirements.WebHome]] **3. Quality Gates Validate Approach**
213 * 2 gates prove the concept
214 * Remaining 5 gates added in POC2
215 * Gates must demonstrably improve quality **4. Iterate Based on Results**
216 * POC results determine next steps
217 * Decision gate after POC1
218 * Flexibility to pivot if needed === 9.2 Success = Clear Path Forward === POC succeeds if we can confidently answer: ✅ **Technical Feasibility:**
219 * Can AI extract claims reliably?
220 * Can AI find balanced evidence?
221 * Can AI compute reasonable verdicts? ✅ **Quality Approach:**
222 * Do quality gates improve output?
223 * Can we measure and track quality?
224 * Is the gate approach scalable? ✅ **Production Path:**
225 * Is the core architecture sound?
226 * What needs improvement for production?
227 * Is POC2 the right next step? == 10. Related Pages == * **[[Main Requirements>>FactHarbor.Specification.Requirements.WebHome]]** - Full system requirements (this POC implements a subset)
228 * **[[POC1 Specification (Detailed)>>FactHarbor.Specification.POC.Specification]]** - Detailed POC1 technical specs
229 * **[[POC Summary>>FactHarbor.Specification.POC.Summary]]** - High-level POC overview
230 * **[[Implementation Roadmap>>FactHarbor.Roadmap.WebHome]]** - POC1, POC2, Beta 0, V1.0 phases
231 * **[[User Needs>>FactHarbor.Specification.Requirements.User Needs.WebHome]]** - What users need (drives requirements) **Document Owner:** Technical Team **Review Frequency:** After each POC iteration **Version History:**
232 * v1.0 - Initial POC requirements
233 * v2.0 - Updated after specification cross-check
234 * v3.0 - Aligned with Main Requirements (FR/NFR IDs added)