Data & Analysis

Gricea automatically logs every meaningful interaction as an event. Each event has a timestamp, event type (kind), and a payload with event-specific data. Here's what gets captured:

Survey Responses — Every survey submission with all question answers, variable names, and timestamps.

Chat Messages — Every message sent by the participant and every response generated by the AI, including the full message text and metadata.

LLM Calls & Responses — Complete records of every LLM API call: the model used, system prompt, user prompt, temperature, tokens consumed, response text, and latency.

RAG Retrievals — What queries were sent to the vector store, what documents were retrieved, relevance scores, and any filters applied (topic, bias mode, metadata).

Variable Changes — Every time a study or task variable is set or modified, including the old value, new value, and which node made the change.

User Interactions — Behavioral signals including button clicks, navigation events, time spent on each page/node, and component interactions.

Timing Data — Timestamps for entering and leaving each node, total time per stage, and session duration.

Flow Progression — The exact path each participant took through the study, including which branch was taken at each decision point and which condition was assigned.

After participants complete your study, you have access to several layers of data:

Episode DataStore — The final state of all variables for each participant session. This is a key-value map containing survey responses, assigned conditions, computed scores, and any other variables set during the study.

Event Logs — The complete chronological event stream for each session. Each event includes:

• episodeId — Which participant session
• kind — Event type identifier
• payload — Event-specific data (responses, messages, API calls, etc.)
• timestamp — When it occurred
• studyVersionId — Which study version

Analytics Dashboard — The study analytics tab provides:

• Completion rates and drop-off points
• Participant status overview (in progress / completed / terminated)
• Survey response summaries
• Session timeline visualization

What you can do with the data:

• Export event logs for offline analysis in Python, R, or your preferred tool
• Query the dataStore for quick variable summaries
• Replay individual sessions step-by-step
• Compare conditions using the assigned variables from randomization
• Analyze chat transcripts with the full context of what the AI showed