Integration Overview
The Context Graph integrates with the broader MATIH platform through Kafka event streaming, orchestrator hooks for non-invasive agent capture, and a metrics bridge that links observability data to thinking traces. These integration points enable the Context Graph to capture agent behavior without modifying core orchestrator logic.
Integration Architecture
+---------------------+ +---------------------+
| BI Orchestrator | --> | Orchestrator Hooks |
+---------------------+ +----------+----------+
|
+---------------------+ +-----------v---------+
| General Orchestrator| --> | Thinking Capture |
+---------------------+ | Service |
+----------+----------+
|
+----------------+--------+--------+
| | |
+--------v-------+ +-----v------+ +--------v--------+
| Kafka Producer | | Dgraph | | Metrics Bridge |
| / Consumer | | Store | | |
+--------+-------+ +------------+ +--------+--------+
| |
+--------v-------+ +--------v--------+
| Kafka Topics | | Observability |
| (streaming) | | (Prometheus) |
+----------------+ +-----------------+Subsections
| Page | Description |
|---|---|
| Kafka Producers and Consumers | Kafka-based event streaming for thinking traces and decisions |
| Orchestrator Hooks | Non-invasive middleware for capturing agent thinking |
| Metrics Bridge | Linking observability spans and metrics to thinking traces |
Design Principles
- Non-invasive -- Integration points do not modify core orchestrator logic
- Graceful degradation -- All hooks are safe to call when services are unavailable
- Tenant-scoped -- All events are tagged with tenant context
- Feature-flagged -- Each integration can be independently enabled per tenant