Skip to content

Cog Arch Actr

cog_arch_actr — ACT-R cognitive architecture.

Cluster: Experience & Autonomy | Type: component | MCP Tools: 34

Overview

Pilot-ready infrastructure, not a polished end-user workflow

cog_arch_actr is functional for MVP and pilot use as an MCP-accessible reasoning component. It is best treated as infrastructure for developers or advanced agent workflows, not as a standalone tool for non-technical users.

The component exposes ACT-R concepts directly: chunks, productions, buffers, activation, and cycles. A first-time vibe coder will usually need a higher-level recipe or wrapper around these operations to get immediate value. Retrieval also includes stochastic ACT-R behavior by design, so exact runs may vary unless callers control thresholds, timing, and random state at the engine layer.

Adaptive Control of Thought-Rational (ACT-R) cognitive architecture implementation that models human cognition through a declarative memory of typed chunks and a procedural production system. Delegates all 26 core operations to an inner ACTRMCPBlock backed by SQLite, with six read-only discovery operations for capabilities, backend status, engine config, store status, schema, and degradation contract.

Discovery is path-redacted and honest about the component's deterministic envelope: stochastic retrieval/utility noise is surfaced, optional LLM authoring is disclosed as peripheral and fallback-capable, and underused persistence tables are marked as live_workflow, audit_only, or infrastructure_ready.

When to use:

  • Simulating human-like associative recall and production-rule reasoning in an agent loop
  • Storing and retrieving typed symbolic chunks to model working memory and long-term declarative knowledge
  • Augmenting ACT-R symbolic reasoning with LLM-generated productions or explanations
  • Inspecting ACT-R capability/status metadata before selecting mutable operations

Do not use as:

  • A general natural-language chatbot or turnkey workflow builder
  • A non-developer-facing onboarding path without a simpler guide or prebuilt workflow
  • A regulated-domain decision system without external validation, audit controls, and human review
  • An arbitrary SQL, filesystem, endpoint-probing, or agentic-runtime adapter

Reliability envelope: Outputs include completion_state, warning_card, evidence, request_id, task_id, and run_id where supplied. Expected domain misses are qualified-draft; degraded or blocked execution is surfaced as blocked-escalated with stable G6_E_* codes.

Example:

from mvp.cog_arch_actr import CogArchACTRBlock, ACTRInput

block = CogArchACTRBlock(name="actr")
create = block.infer(ACTRInput(
    op="chunk_create",
    chunk_type="fact",
    slots_json='{"subject":"sky","color":"blue"}',
    current_time=0.0,
))
result = block.infer(ACTRInput(
    op="chunk_retrieve",
    chunk_type="fact",
    slots_json='{"subject":"sky"}',
    current_time=1.0,
))
# result.ok -> True; result.value -> ACTROutput with records, chunk_id, message

Works well with: cog_arch_soar, experience_loop, deep_understanding

Public API

CogArchACTRBlock(AIBlock[ACTRInput, ACTROutput, dict])

ACT-R cognitive architecture block — delegates to ACTRMCPBlock.

Field Type Default
name str 'cog_arch_actr'
state dict \| None None
db_path str ':memory:'
resource_bounds ResourceBounds \| None None
usage ResourceUsage field(default_factory=ResourceUsage)

Methods:

infer(data: ACTRInput) -> Result[ACTROutput]

ACTRInput(BaseModel)

Field Type Default
op str required
chunk_id str ''
chunk_type str ''
slots_json str '{}'
production_id str ''
production_name str ''
conditions_json str '{}'
actions_json str '{}'
utility float 0.0
buffer_name str ''
max_cycles int 100
goal_test_json str ''
current_time float 0.0
decay float 0.5
reward float 0.0
alpha float 0.2
description str ''
buffer_state_json str ''
source_json str ''
target_json str ''
bridge_config_json str ''
gate_type str ''
query str ''
top_k int 5
limit int 50
session_id str ''
request_id str ''
task_id str ''
run_id str ''

ACTROutput(BaseModel)

Field Type Default
op str required
ok bool required
message str ''
error str ''
data_json str ''
chunk_id str ''
production_id str ''
chunks list[dict] Field(default_factory=list)
productions list[dict] Field(default_factory=list)
records list[dict[str, Any]] Field(default_factory=list)
count int 0
metadata dict[str, Any] Field(default_factory=dict)
degraded bool False
degradation_reason str ''
completion_state str 'qualified-draft'
warning_card dict[str, Any] Field(default_factory=dict)
evidence dict[str, Any] Field(default_factory=dict)
code str ''
request_id str ''
task_id str ''
run_id str ''

MCP Tools

Operation Source
ops cog_arch_actr_mcp
help cog_arch_actr_mcp
actr_capabilities cog_arch_actr_mcp
actr_backend_status cog_arch_actr_mcp
actr_engine_config cog_arch_actr_mcp
actr_store_status cog_arch_actr_mcp
actr_schema_describe cog_arch_actr_mcp
actr_degradation_contract cog_arch_actr_mcp
chunk_create cog_arch_actr_mcp
chunk_get cog_arch_actr_mcp
chunk_retrieve cog_arch_actr_mcp
chunk_list cog_arch_actr_mcp
chunk_delete cog_arch_actr_mcp
production_create cog_arch_actr_mcp
production_get cog_arch_actr_mcp
production_list cog_arch_actr_mcp
production_delete cog_arch_actr_mcp
buffer_set cog_arch_actr_mcp
buffer_get cog_arch_actr_mcp
buffer_clear cog_arch_actr_mcp
cycle_run cog_arch_actr_mcp
cycle_step cog_arch_actr_mcp
cycle_status cog_arch_actr_mcp
activation_compute cog_arch_actr_mcp
utility_update cog_arch_actr_mcp
utility_list cog_arch_actr_mcp
llm_encode_chunk cog_arch_actr_mcp
llm_suggest_production cog_arch_actr_mcp
find_analogy cog_arch_actr_mcp
realtime_bridge cog_arch_actr_mcp
hat_gate cog_arch_actr_mcp
actr_search cog_arch_actr_mcp
actr_info cog_arch_actr_mcp
list_patterns cog_arch_actr_mcp