Chapter 18: Cognitive Operators (Trace, Gavel, Bridge)

The Bridge Operator: Syntax for high-density extraction

The Bridge is the atomic operator of extraction. It connects a known state (Context) to an unknown state (Signal).

• Syntax: Bridge(prefix: string, slot: Slot, suffix: string)
• Visual Representation: A connector cable linking a document node to a data field node.

The Bridge operator enforces Temperature 0.1. It is designed for “Fact Retrieval,” not creativity. In the IDE, dropping a “Bridge Node” onto the canvas automatically configures the underlying model parameters for maximum determinism.

[IMAGE PROMPT: A UI mock-up of the ‘Bridge Node’. It has an input port (‘Context’) and an output port (‘Signal’). Inside the node, a visualization shows raw text entering a funnel and compressing into a single data point. A ‘Density Meter’ gauge shows ‘98% Subtractive Efficiency’.]

The Gavel Operator: Syntax for validation and quality gates

The Gavel is the boolean logic gate of the cognitive circuit.

• Syntax: Gavel(content: string, criteria: string) -> boolean
• Function: It creates a “Branch” in the control flow.

If Gavel == Pass, the data flows downstream. If Gavel == Fail, the data flows into a “Retry Loop” or an “Error Handler.”

[IMAGE PROMPT: A flow-chart view of a ‘Quality Gate’. A ‘Gavel Node’ sits between an ‘Extraction’ and a ‘Database Write’. The Gavel Node has a green output wire (‘Pass’) and a red output wire (‘Fail’). The red wire loops back to the input, visualizing a self-correction cycle.]

The Detective Operator: Compound syntax for Reasoning + Extraction

The Detective is a macro-operator. It encapsulates the “Reasoning Dredge” pattern we discovered in the lab.

• Syntax: Detective(context, question)
• Expands to: [TRACE_ANCHOR, {slot: trace}, RESULT_ANCHOR, {slot: result}]

This operator hides the complexity of “Thinking Tokens” from the developer. You simply drop a “Detective Node” when you need to solve a contradiction. The Compiler automatically manages the token budget, ensuring enough space is reserved for the trace before the result is generated.

Standard Library of Operators

A robust programming language needs a stdlib. The Dredge Standard Library provides pre-compiled operators for common tasks:

• redact_pii(text) -> Returns text with names/phones replaced by [REDACTED].
• normalize_date(text) -> Returns ISO 8601 string.
• sentiment_check(text) -> Returns -1.0 to 1.0 float.

These are not code functions; they are Cognitive Functions. They don’t run on the CPU; they run on the NPU (Neural Processing Unit). But to the developer, they look and feel exactly like calling a standard library function.

[IMAGE PROMPT: A ‘Library Palette’ on the left side of the IDE. Categories include ‘Extraction’, ‘Logic’, ‘Compliance’, ‘Transformation’. The user is dragging a ‘PII Redactor’ operator from the palette onto the canvas to wire it into a document processing pipeline.]