Memory System
Why memory matters
Standard chatbots forget everything when the conversation ends. Ask the same question twice and you get the same answer — no context, no learning, no continuity.
AI Partner uses a 5-layer memory system that persists across conversations, days, and weeks. When you ask about a vendor you spoke to last month, AI Partner pulls up their record — their communication style, open commitments, last contact date — without you repeating it.
The 5 layers
1. Episodic Memory — the event timeline
What it stores: A timestamped record of everything that happens — goals run, emails sent, meetings attended, files created, decisions made.
Think of it as: A diary. Every event is stored with its type, timestamp, content, and metadata.
Example entries:
2026-05-10 09:15 goal_completed "Researched NVIDIA Q1 earnings; created slide deck"
2026-05-10 10:30 email_sent "Replied to sarah@acme.com re: proposal"
2026-05-10 14:00 meeting_attended "Joined Teams call with Sequoia; 45 min; action items extracted"
2026-05-11 07:05 heartbeat_task "Morning briefing: 5 news items delivered to Telegram"
Queryable by: time range, event type, keyword, or semantic similarity.
2. Biographic Facts — semantic knowledge about you
What it stores: Subject / Predicate / Object facts consolidated from the episodic timeline. Facts have confidence scores and are updated over time.
Think of it as: A knowledge graph of things the agent knows to be true about you and your world.
Example facts:
(Alex, works_at, Acme Inc.) confidence: 1.0
(Alex, is_raising, Series A) confidence: 0.95
(Sequoia, is_contact_of, Alex) confidence: 0.9
(Alex, prefers_communication_style, direct+data) confidence: 0.85
(Acme v2.0, launches_by, 2026-07-15) confidence: 0.8
Built automatically from what you tell the agent and what it observes. No manual tagging required.
3. Counterparty Store — your contact graph
What it stores: One stable record per person you interact with, linked across every channel.
Bob at Acme might be bjones@acme.com in email, @bjones in Slack, and user 123456789 in Telegram. The counterparty store unifies all three into one record:
Name: Bob Jones
Company: Acme Inc.
Class: client
Aliases:
- bjones@acme.com (email)
- @bjones (Slack)
- 123456789 (Telegram)
Tone: formal-friendly; responds quickly; prefers bullet points
Last contact: 2026-05-08 via email
Open commitments:
- "Will send revised proposal by May 15" (from meeting on May 8)
Used by: the authority policy (gating by relationship class), email/DM proxy (personalising reply tone), meeting proxy (recognising who's speaking).
4. Search by meaning
What it does: Finds things by what they mean, not by the words you happened to use.
Think of it as: Asking a colleague with a good memory. "Find everything related to our funding round" surfaces the relevant conversations even where nobody wrote the phrase "funding round".
It also avoids the failure mode that makes search useless in practice — returning ten near-identical results. What comes back is relevant and varied.
Used automatically whenever the agent is working something out. You never have to invoke it.
5. Your documents — the Knowledge Base
What it holds: The documents you give it — handbooks, specs, contracts, decks.
Think of it as: A searchable library of your own material. Upload your company handbook, investor deck or technical spec, and the agent can quote the relevant part when it matters.
How to use:
- Go to Knowledge Base in the sidebar
- Add a document
- Wait a moment while it's indexed
- Ask: "Based on our investor deck, what is our go-to-market strategy?"
It searches your documents by meaning and by exact wording, and names the source it answered from.
How the agent uses memory
Every time AI Partner works out what to do next, it searches its own memory for anything relevant to what's in front of it, and takes it into account before deciding.
You never have to say "remember when…". It happens on every step, whether you're in a quick chat, a long goal, or a meeting it's attending as you.
Querying memory yourself
You can ask the agent to surface memories directly:
"What do I know about Bob Jones?"
→ Returns counterparty record + recent episodic events involving Bob
"What happened in last week's meetings?"
→ Returns episodic events of type meeting_attended from the last 7 days
"What have I committed to recently?"
→ Returns open commitments from counterparty store
"Search my knowledge base for our refund policy"
→ Runs hybrid search across uploaded documents
Or use the Memory Inspector panel (sidebar → Memory) to browse all 5 layers visually.
Privacy and persistence
Memory stays on your own instance. It isn't sent anywhere else, isn't pooled with other users, and isn't used to train anything.
- Review it — the Memory Inspector shows every layer, so nothing it knows about you is hidden from you.
- Correct it — tell it in conversation ("my title is X now, not Y") and the correction sticks.
- Delete it — remove individual entries from the Memory Inspector.
On a multi-user instance, each person's memory is theirs alone: it's isolated by login, and administrators can't read it.
Your documents
The fifth layer is the one you curate deliberately, and it has its own guide: adding documents, searching them yourself, and organisation-wide sources an administrator can connect.
Search quality
Searching by meaning — so that "time off policy" finds a passage about annual leave — needs a model that specialises in it. AI Partner uses the best one you've connected and falls back on its own if none is available.
| What you've connected | What search does |
|---|---|
| OpenAI or Cohere | Full meaning-based search — the recommended setup |
| A local model | Meaning-based search, running entirely on your own machine |
| Nothing extra | Keyword search — always works, but matches words rather than meaning |
Connect one of the first two in Settings → Models if you rely on your Knowledge Base for answers; keyword-only search will miss a passage that means the right thing in different words.