docProtocols
Protocols are the digital sensors of the Blankstate ecosystem—deterministic instruments for measuring human interaction. They define what to observe, how to measure it, and why it matters.
Every protocol must satisfy these invariants. Any protocol that violates these constraints is not a Blankstate protocol.
Same input + protocol = Same output. Always.
No training required per use case.
Protocol defines its own measurement criteria.
Every finding traceable to source.
Sub-second analysis for streaming.
Can run locally, federated, or cloud.
How Protocols Work
1. Define What to Observe
Protocols contain nuances—specific patterns, concepts, or behaviors to detect. Each nuance has semantic indicators and scoring weights.
2. Process Through the Engine
The IBF Engine ingests interaction data and converts it to a Unified Interaction Representation (UIR) before protocol analysis. No raw content is stored—only structured signals.
3. Produce Traceable Results
Every score, every detection, every insight links back to the source. You can always ask "why?" and get an evidence-based answer. That's the glass-box guarantee.
Interaction Types
Protocols measure interactions across all relationship types.
Conversations, meetings, negotiations, interviews, written exchanges
Voice assistants, chatbots, UI interactions, forms, navigation
API calls, event streams, log sequences, data pipelines
Sensor data, biometrics, IoT interactions, physical-digital bridging
Universal Modality Coverage
Protocols are modality-agnostic. All modalities convert to a Unified Interaction Representation (UIR) before protocol analysis. One protocol. Any modality. Same measurement.
| Status | Modality | Data Types | Notes |
|---|---|---|---|
| ✓ | Text Documents | .pdf, .docx, .pptx, .rtf, .txt | Full extraction with page/slide segmentation |
| ✓ | Structured Data | .xlsx, .csv, JSON, XML | Sheet/table parsing with structure preservation |
| ✓ | Scanned Documents | Image-based PDFs, photos of documents | OCR extraction via advanced handler |
| ✓ | Audio | Calls, meetings, voice notes, podcasts | Transcription with speaker diarization and timing |
| ◐ | Images | Embedded images, screenshots, diagrams | VLM description; OCR for text-in-image |
| ○ | Video | Screen recordings, video meetings | Frame extraction, visual context |
| ◐ | System Logs | Application logs, error traces, audit trails | Text-based logs supported; event semanticization planned |
| ○ | Event Streams | API request/response, webhooks, pub/sub | Real-time ingestion pipeline |
| ○ | Behavioral Signals | Click streams, navigation paths, form submissions | Session reconstruction |
| ○ | Sensor Data | IoT telemetry, biometrics, environmental | Signal abstraction |
Abstraction Principle
Regardless of source modality, the protocol engine receives:
Current Release
Document + audio transcription with entity extraction