Documentation / Concepts

Query Model

The conceptual query model moves from language expression through CongSynth reasoning into replayable decision output.

Congregation DB is queried as a decision graph rather than as a flat table or isolated vector surface. The query model is designed to preserve context while reasoning moves across entities, relationships and heterogeneous signals.

Querying a Decision Graph

Querying a decision graph means reasoning over connected context. A query may move across entities, dependency paths, truth lanes and evolving signal states rather than retrieving a single disconnected record.

The model is intended for decision environments where provenance and interpretability matter as much as the output itself.

Conceptual Query Flow

Query pipeline

CongLang Query

A conceptual reasoning expression describes what should be evaluated across the decision environment.

CongIR Representation

The expression is preserved in an intermediate reasoning form before CongSynth execution occurs.

CongSynth Traversal

Relevant entities, relationships and dependency paths are traversed through the context synthesis engine.

Signal Evaluation

Signals are evaluated with provenance, truth-lane distinctions and surrounding context still attached.

Decision Output

The resulting output remains tied to the reasoning path that produced it.

A CongLang query moves through CongIR, traverses CongSynth, evaluates signals, and produces an output that remains tied to the reasoning path behind it.

Query Expression

At the query-expression layer, reasoning intent is described without binding that intent to a specific execution mechanism. This allows the model to remain conceptual and stable while the system preserves provenance and contextual structure underneath.

CongLang
MATCH portfolio_company
WHERE covenant_status = "active"

RELATE supplier
RELATE logistics_route
WHERE disruption_probability > 0.71

RETURN continuity_risk
WITH provenance

In plain language, this conceptual query starts from active portfolio companies, follows supplier and logistics-route context, evaluates disruption risk, and returns a continuity-risk view with provenance retained.

Intermediate Representation

Intermediate reasoning layer

CongIR preserves reasoning semantics between language expression and CongSynth execution.

The intermediate representation separates how a query is expressed from how reasoning is carried forward through the system. This keeps the model stable at the semantic layer without exposing execution detail.

CongSynth Reasoning

Once represented conceptually, the query traverses CongSynth across relevant entities, relationships and dependency paths. This is where surrounding context remains attached instead of being flattened away before evaluation.

This is also where truth-lane distinctions remain useful. The system can reason over deterministic and probabilistic signals without collapsing them into a single undifferentiated result.

Decision Output

A decision output is more than a returned value. It is a contextual outcome shaped by a specific reasoning path, signal environment and graph traversal.

The conceptual goal is not simply to answer a query, but to preserve enough structure that the resulting output remains inspectable and meaningful.

Decision Replay

Decision replay means a system can reconstruct how a specific outcome was formed from the signals and graph paths that contributed to it.

Replay trace
decision_output: continuity_risk = elevated
query_model:
  language: CongLang
  representation: CongIR
reasoning_path:
  portfolio_company -> supplier -> logistics_route
signals_used:
  - covenant_status (deterministic)
  - disruption_probability (probabilistic)
traceability: replayable