What am I looking at?
Each time the investigator agent finishes a work session, it saves a complete
copy of its memory — every hypothesis, finding, piece of evidence, and
suspect score — from CockroachDB to Amazon S3. Each of those saved copies
is one
frame below.
Drag the slider to move backward and forward
through the investigation's history and watch the agent's thinking evolve:
- The counters show how much the agent had discovered by that point in
time (hypotheses it was pursuing, findings it recorded, evidence it collected).
- Each bar is one suspect; its length and number are that person's
guilt score at that moment. Watch scores rise as evidence accumulates.
- A green-outlined row is a suspect who first
appears in this frame — a new lead the agent just uncovered.
- Hover a row to read the agent's reasoning recorded at that time.
This is only possible because the agent's memory
persists in
CockroachDB. A stateless chatbot would have nothing to replay.
Click the tiles above to switch
between the agent's hypotheses, findings, and the evidence it collected by this
point in the investigation.