| a81193bc87a02963… | Sets are named many-to-many CXU groupings, and the UNIVERSAL set is always in scope so that CXUs can be organized into reusable scoped collections. | active | contextual | yes | |
| 3ed56255adff07d1… | Rank 0 CXUs are set-scoped CXUs that are always included, determined by `is_rank0` on set membership, so that certain context units are guaranteed to be in scope. | active | contextual | yes | |
| 53f660b96361aa1b… | The CXU lifecycle progresses through DRAFT, ACTIVE, INACTIVE, and VERSIONED states, with an approval workflow required for REGULATED and ABSOLUTE classifications so that controlled knowledge changes can be governed. | active | contextual | yes | |
| b16ebe37a41fe4dd… | A CXU, or Context Unit, is an atomic knowledge unit containing a claim, supporting quotes, and metadata, and it uses a content-addressable ID formed from 1220 plus SHA-256 so that each unit is self-contained and addressable. | active | contextual | yes | |
| 23638feb877f25a0… | Lifecycle endpoints provide suggestions, approvals, and lifecycle state transitions so that CXU lifecycle management actions can be performed through the API. | active | contextual | yes | |
| 7c5496fa8f91c387… | CORTEX endpoints provide status, agent health, quality scores, knowledge gaps, ambient research, and metrics so that operational and analytical information about agents can be queried. | active | contextual | yes | |
| 447714e925c71231… | The FastAPI API surface exposes routes including `/cxu`, health endpoints, context endpoints, rank0 endpoints, sets endpoints, cortex endpoints, reports endpoints, and document-analysis endpoints so that multiple platform capabilities are available over HTTP. | active | contextual | yes | |
| ea59544ec3d4ab40… | The `context_engine.kag` package contains the Neo4j graph backend, Qdrant vector backend, and embedding generator so that graph, vector, and embedding functions are implemented together. | active | contextual | yes | |
| 8a318705a60c8d26… | The `context_engine.services` package contains the query processor with a three-strategy pipeline and the Text2Cypher service so that query execution and Cypher generation services are provided there. | active | contextual | yes | |
| 33b8023b6cdc17fd… | The `context_engine.cortex` package contains CORTEX agents, the message bus, TEMPR memory, reranker, and set-relationship registry so that agent orchestration and memory functions are grouped in one package. | active | contextual | yes | |
| 0b8fab58f6565ac4… | The `context_engine.store` package provides the PostgreSQL storage adapter and hybrid search engine so that persistence and mixed retrieval capabilities are implemented there. | active | contextual | yes | |
| b64a287fbe152d87… | The `context_engine.extract` package performs LLM extraction through a LiteLLM proxy and Instructor so that structured output can be produced from extraction workflows. | active | contextual | yes | |
| a108dd432da0fcde… | The workspace structure includes three Python packages, and `context_engine.core` contains CXU models, consolidation, and similarity detection so that core knowledge-unit logic is centralized there. | active | contextual | yes | |
| c262bb21c2ff1d2e… | The extraction pipeline accesses LLMs through a LiteLLM proxy that is OpenAI-compatible so that model routing is handled through a compatible proxy layer. | active | contextual | yes | |
| 4dbfc668c342c49e… | NATS provides L2 agent coordination in the Context Engine and uses nats-py as the messaging client. | active | contextual | yes | |
| 70bccc293e832f51… | Redis serves as the cache backend in the Context Engine and uses the redis client library for access. | active | contextual | yes | |
| 307084656ec4b880… | Qdrant provides vector search with 768-dimensional embeddings in the Context Engine, using qdrant-client as the integration technology. | active | contextual | yes | |
| 842cd5d3561377f6… | Neo4j provides the knowledge graph and TEMPR memory backend in the Context Engine, using neo4j-driver for connectivity. | active | contextual | yes | |
| b16f27dce79e2289… | PostgreSQL serves as the source of truth and provides JSONB and full-text search capabilities in the Context Engine, using asyncpg as its technology adapter. | active | contextual | yes | |
| 51062681f36fa05b… | The Context Engine is a multi-backend Knowledge-Augmented Graph system with five storage and messaging backends so that knowledge, vectors, cache, graph, and coordination functions are distributed across specialized services. | active | contextual | yes | |
| 8e62c2eb45642d8b… | Client overlays do not contain platform code, and instead only configure tenant-specific behavior so that shared platform implementation remains outside tenant directories. | active | contextual | yes | |
| 1aaaaac69ae42c48… | Per-tenant overlays contain agent definitions, workflow configs, data connectors, and SQL templates so that platform behavior can be configured separately for each tenant. | active | contextual | yes | |
| dea24707f8ddf5dd… | Multi-tenancy is implemented through client overlays in the `clients/` directory, such as `clients/jnj/` and `clients/patrick/`, so that tenant-specific configuration is isolated by client. | active | contextual | yes | |
| 2843392f2f4269ff… | The platform uses GitHub Actions for CI/CD and changesets for versioning so that build, deployment, and release version workflows are managed through those tools. | active | contextual | yes | |
| 35e73b7f638addf4… | The platform uses uv for Python package management and pnpm for TypeScript package management so that dependency management is separated by language ecosystem. | active | contextual | yes | |
| 65cbc2566ebb2086… | The `IN_SET` edge represents CxU to Set membership so that graph queries can determine which context units belong to which named sets. | active | contextual | yes | |
| 0707f2a5c5e5d7cc… | The `DERIVED_FROM` edge connects one CxU to another CxU for lineage so that derivation relationships between context units are preserved in the graph. | active | contextual | yes | |
| 624852e08516d466… | The `PRODUCED` edge connects an AgentRun to a CxU so that the graph records which agent execution generated a given context unit. | active | contextual | yes | |
| 1aec7d659a1bad64… | The `BELONGS_TO` edge connects a child node to a parent node in the hierarchy so that hierarchical organizational relationships are represented explicitly in the graph. | active | contextual | yes | |
| 2c5ba3ae58a11457… | The `CONTEXT_FOR` edge connects a CxU to an entity and carries the properties dimension, q_value, and is_rank0 so that contextual relevance and ranking metadata are stored on the relationship. | active | contextual | yes | |
| ea84d64265432a32… | The Neo4j graph schema defines memory nodes as MemoryNode with relationships TEMPORAL_NEXT, SEMANTIC_SIMILAR, and REFERENCED so that memory structures can encode temporal, semantic, and reference links. | active | contextual | yes | |
| 9feea05abe525971… | The Neo4j graph schema defines agent nodes named AgentRun, ActionItem, and Analysis so that agent activities and outputs can be represented as graph entities. | active | contextual | yes | |
| 30238db26dcf22d4… | The Neo4j graph schema defines hierarchy nodes for Company, Pillar, Group, BusinessUnit, VirtualBusinessUnit, and CrossUnit so that organizational structure can be represented in the graph. | active | contextual | yes | |
| 843d6776d109ea2b… | In the DraftCXU data model, `confidence` is a float ranging from 0.0 to 1.0 so that extraction certainty can be represented numerically. | active | contextual | yes | |
| eeb694047694488b… | In the DraftCXU data model, `keywords` is a list so that searchable or descriptive terms can be attached to each context unit. | active | contextual | yes | |
| 48328b6676fbf8c7… | In the DraftCXU data model, `parent_assumption_ids` is a list so that assumptions can reference parent assumptions for lineage or dependency tracking. | active | contextual | yes | |
| 401426a3bfd48fde… | In the DraftCXU data model, `is_assumption` is a boolean field so that a context unit can explicitly indicate whether it represents an assumption. | active | contextual | yes | |
| e3a4d039efb55351… | In the DraftCXU data model, `claim_type` is an enum that includes values such as definition, requirement, and procedure so that the form of each claim can be classified. | active | contextual | yes | |
| 7a5f1cf119789e2f… | In the DraftCXU data model, `knowledge_type` is an enum with values `axiom`, `derived`, or `prescribed` so that each context unit is categorized by its knowledge status. | active | contextual | yes | |
| 6deaed392c955b54… | In the DraftCXU data model, `supporting_quotes` is a list containing 3 to 6 verbatim source quotes so that each claim is backed by direct evidence from the source material. | active | contextual | yes | |
| 3441aa1764302796… | In the DraftCXU data model, `claim` is a standalone assertion between 15 and 70 words so that each context unit expresses a self-contained statement at a controlled length. | active | contextual | yes | |
| 6e9ec08f3901d97d… | In the DraftCXU data model, `cxu_id` is a string formatted as `1220` plus a SHA-256 hex digest with 68 characters so that each draft context unit has a deterministic identifier format. | active | contextual | yes | |
| d102ce8f0f1b5282… | Named relationships are managed through a set-relationship registry with a lifecycle of pending, active, and inactive states, plus usage tracking and co-occurrence discovery so that relationship definitions can be governed and analyzed over time. | active | contextual | yes | |
| 501ddc2ba5a89192… | Extraction is a two-phase process of chunking followed by parallel lens extraction, using lenses for facts, definitions, procedures, requirements, and specifications, with profiles for general and pharma-manufacturing domains so that extraction can be tailored by content type and domain. | active | contextual | yes | |
| 91a44bde822d7bc9… | The CORTEX agent layer includes an AmbientAgent for proactive gap detection so that missing knowledge or coverage gaps can be identified without explicit user prompting. | active | contextual | yes | |
| 0e1f65d42dac8fcb… | The CORTEX agent layer includes a LearningAgent that uses TEMPR memory so that learned or retained contextual information can be managed through the agent system. | active | contextual | yes | |
| 00a47d653bd14e94… | The CORTEX agent layer is coordinated and includes a RetrievalAgent for Q-value reranking, cross-encoder, and two-level retrieval so that retrieval quality can be improved through layered ranking logic. | active | contextual | yes | |
| 5fe15bbf179e18dd… | KAG retrieval uses a three-strategy pipeline consisting of template Cypher, edge-filter Cypher, and Text2Cypher free search, and the results are fused with Q-value weighted RRF so that multiple retrieval methods contribute to final ranking. | active | contextual | yes | |
| 37f3e4b4ea785f34… | Sets are named many-to-many CXU groupings, and the UNIVERSAL set is always in scope so that context units can be grouped flexibly while preserving a default global scope. | active | contextual | yes | |
| cbd72b6477c3c280… | Rank 0 CXUs are set-scoped CXUs that are always included, indicated by `is_rank0` on set membership, so that certain context units remain in scope by default within a set. | active | contextual | yes | |