Althea initializes every user relationship by constructing a Dossier. This is a structured semantic profile derived from public data that allows the agent to understand who you are, what you work on, and what you know before the first interaction occurs.
Composite Search Pipeline
At signup, the system executes a parallel retrieval process to aggregate a "Evidence List" from three distinct layers of the web.
- Professional Data: Queries to structured databases establish current role, employment history, and industry.
- Semantic Search: Queries to neural search engines identify authored content, technical talks, and open source contributions.
- General Index: Queries to standard search engines fill biographical gaps.
An LLM synthesis step fuses these fragmented data points into a coherent narrative, resolving ambiguities and formatting the output for machine consumption.
Structured Memory
The Dossier is not an unstructured blob of text. It is persisted as a JSON schema divided into three active memory banks:
- Experience: A timeline of professional roles and organizational affiliations.
- Expertise: A validated graph of technical skills and domain knowledge.
- Interests: A map of stated preferences and research focus areas.
Latency and Utility
This entire pipeline executes during the onboarding flow. When the user lands on the chat interface, the agent has already loaded this context into its system prompt.
This enables immediate, high-fidelity interaction. If you ask a technical question, Althea checks your Expertise to calibrate the complexity of the response. If you request news, she filters against your Interests.
The Dossier is a mutable object. Users retain full read/write access to this data in their settings, ensuring the agent's model of the user remains accurate and transparent.