Skip to Content
Last modified by Robert Schaub on 2026/02/08 08:31
Show last authors
1 {{warning}}
2 **Experimental (v2.6.33)** - Flexible output structure not bound to canonical schema. File: ##monolithic-dynamic.ts## (~550 lines). Labeled as "Experimental" in UI.
3 {{/warning}}
4
5 = Monolithic Dynamic Pipeline Internal Flow =
6
7 {{mermaid}}
8
9 flowchart TB
10 subgraph Entry[Entry Point]
11 INPUT[runMonolithicDynamic]
12 BUDGET[Initialize Budget maxIterations 4 maxSearches 6]
13 TIMEOUT[Start timeout 2.5 minutes]
14 end
15
16 subgraph Analysis[Single LLM Analysis]
17 PROMPT[Build dynamic analysis prompt]
18 LLM[LLM Call DynamicAnalysisSchema]
19 SEARCH[searchWebWithProvider]
20 FETCH[fetchSourceContent]
21 end
22
23 subgraph Safety[Safety Contract Enforcement]
24 CITATIONS[Extract citations array]
25 RAW[Store rawJson]
26 VALIDATE[Validate minimum fields]
27 end
28
29 subgraph Transform[Output Transform]
30 DYNAMIC[Build DynamicAnalysisResult]
31 REPORT[Generate experimental report]
32 end
33
34 subgraph Output[Output]
35 RESULT[resultJson dynamic structure]
36 MARKDOWN[reportMarkdown experimental label]
37 end
38
39 INPUT --> BUDGET
40 BUDGET --> TIMEOUT
41 TIMEOUT --> PROMPT
42
43 PROMPT --> LLM
44 LLM --> SEARCH
45 SEARCH --> FETCH
46 FETCH --> LLM
47
48 LLM --> CITATIONS
49 CITATIONS --> RAW
50 RAW --> VALIDATE
51
52 VALIDATE --> DYNAMIC
53 DYNAMIC --> REPORT
54 REPORT --> RESULT
55 REPORT --> MARKDOWN
56
57 {{/mermaid}}
58
59 = Key Characteristics =
60
61 |= Feature |= Description
62 | **Flexible Output** | LLM can structure analysis freely
63 | **Experimental** | Labeled as experimental in UI
64 | **Safety Contract** | Must always include citations[] and rawJson
65 | **Shorter Budget** | More restrictive limits than canonical
66 | **No Fallback** | Does not fall back to other pipelines
67
68 = Budget Configuration =
69
70 |= Parameter |= Value |= Comparison to Canonical
71 | maxIterations | 4 | vs 5 (20% less)
72 | maxSearches | 6 | vs 8 (25% less)
73 | maxFetches | 8 | vs 10 (20% less)
74 | timeoutMs | 150000 | vs 180000 (2.5 min vs 3 min)
75
76 = Minimum Safety Contract =
77
78 **Required fields (always present):**
79
80 |= Field |= Type |= Purpose
81 | ##citations## | Array | Source URLs with excerpts (provenance)
82 | ##rawJson## | Any | Full LLM output for auditing
83
84 **Optional fields (LLM decides):**
85
86 |= Field |= Type |= Purpose
87 | ##summary## | String | Analysis summary
88 | ##verdict## | Object | label, score, confidence, reasoning
89 | ##findings## | Array | Key findings with support levels
90 | ##methodology## | String | Analysis approach description
91 | ##limitations## | Array | Known analysis limitations
92
93 = Output Schema =
94
95 {{mermaid}}
96
97 erDiagram
98 DYNAMIC_RESULT ||--o{ CITATION : has
99 DYNAMIC_RESULT ||--o| VERDICT : may_have
100 DYNAMIC_RESULT ||--o{ FINDING : may_have
101
102 DYNAMIC_RESULT {
103 array citations_required
104 any rawJson_required
105 string summary
106 string methodology
107 array limitations
108 }
109
110 CITATION {
111 string url_required
112 string title
113 string excerpt_required
114 string accessedAt
115 }
116
117 VERDICT {
118 string label
119 number score_0_to_100
120 number confidence_0_to_100
121 string reasoning
122 }
123
124 FINDING {
125 string point
126 string support_strong_moderate_weak_none
127 array sources
128 string notes
129 }
130
131 {{/mermaid}}
132
133 = Differences from Other Pipelines =
134
135 |= Aspect |= Orchestrated |= Mono Canonical |= Mono Dynamic
136 | Schema | Canonical fixed | Canonical fixed | Flexible
137 | UI | Jobs page | Jobs page | Dynamic viewer
138 | Fallback | Default | To Orchestrated | None
139 | Budget | Generous | Moderate | Restrictive
140 | Label | Standard | Standard | Experimental
141 | Code | ~9000 lines | ~1100 lines | ~550 lines
142
143 = Use Cases =
144
145 **When to use Dynamic:**
146 * Exploratory analysis of novel claim types
147 * Research where canonical structure is limiting
148 * Comparing LLM reasoning approaches
149 * Debugging LLM behavior
150
151 **When NOT to use:**
152 * Production fact-checking
153 * Results that need UI rendering
154 * Comparing results across pipelines