HISTORICAL INTELLIGENCE / OVERVIEW
Research overview
Isolated & healthy
DECISION RESEARCH ENGINE
Reconstruct every decision.
Test every alternative.
Replay Agent69 consultations against exact historical context without touching production execution.
MARKET STATE↓SELECTED AGENTS↓LLM CONSULTATION↓FORWARD REPLAY
CONSULTATION STREAM
Recent historical decisions
DATASET BUILDER
Define an immutable cohort
Loading the authoritative recorder catalog.
VERSIONED DATASETS
Saved cohorts
EVERY LLM CALL
Consultation inspector
Select a consultation to inspect its exact input, response, reasons and chain.
APPEND-ONLY PROMPTS
Create a prompt version
PROMPT LIBRARY
Versions
EXPERIMENT BUILDER
One snapshot. A different intelligence.
Dataset identity is loading.
Feature agents
The complete discovered catalog is shown below. Only checked agent fields enter the model payload; the full raw context remains stored in the immutable snapshot.
Simulation assumptions
Payload preview will prove excluded agents are absent.
BOUNDED RESEARCH DIRECTOR
Hypothesis to shadow evidence
The director searches replay configurations, applies statistical gates, and can only create a shadow challenger. Production eligibility remains false.
Search library: baseline, LONG/SHORT/NO_TRADE parameter variants, and one-feature ablations.
AUTONOMOUS EVIDENCE
Research cycles
Champion / challenger registry
CANDIDATE VS ORIGINAL
Experiment results
RUN COMPARISON
Configuration comparison
Run a second experiment to compare configurations.
ISOLATION & OBSERVABILITY
System health
Production boundary
This service has no exchange client, order-submission method, production database writer, scheduler hook, or bot import. External LLM calls are disabled unless explicitly enabled for this process.