debate.tellodb
Summary
Vector databases retrieve your users' old preferences right alongside their new ones — same confidence, same ranking, zero awareness that one superseded the other. TelloDB is a memory engine built specifically to fix that.
The core mechanism is fact supersession: when a user moves from NYC to SF, TelloDB marks the old location as stale and filters it from active agent context — so the LLM never hallucinates a two-year-old truth. A hybrid HNSW vector plus BM25 search index handles recall, while a separate Metric Vault layer resolves numeric queries deterministically before they ever reach the LLM. The vendor reports p99 retrieval at 4.2ms and benchmarks recall precision above 95% on LongMemEval-S against 68% for standard RAG. The engine ships as a single Rust binary, self-hostable or deployable on the vendor's platform. At v0.1.0, the surface area is narrow — this is a memory layer, not a full agent runtime.
Bottom line: Pick TelloDB when your agent needs to know that a user switched cars last Tuesday and own that answer deterministically; look elsewhere when you need the memory layer bundled inside a full agent orchestration framework.
Community Performance Report Card
No community ratings yet. Be the first to rate this tool!
Community Benchmarks Community
Sign in to submit a benchmarkNo community benchmarks yet. Be the first to share a real-world data point.
Pros
Sign in to edit- Fact supersession automatically marks prior user states as stale when contradicted by new input, so your agent stops confidently telling a user their old address is current.
- Deterministic aggregation in the Metric Vault resolves count and numeric queries before the LLM sees them, which means you stop relying on the model to do arithmetic over memory and stop getting wrong counts.
- Hybrid HNSW vector plus BM25 search runs in a single Rust binary, so you avoid stitching together a vector store and a keyword search service as separate infrastructure dependencies.
- Self-host path with an air-gapped proxy gateway option, so teams with data residency requirements can run the memory layer inside their own perimeter without routing user data through a third-party hosted service.
- Distillation pipeline extracts structured facts from raw conversational text rather than storing full transcripts, which means context windows stay narrow and you are not paying to re-embed every filler word.
Cons
Sign in to edit- TelloDB is a memory substrate only — it provides no agent task planning, tool-calling scaffolding, or workflow logic. Teams that need a full agent runtime will integrate TelloDB as a dependency inside a separate framework (LangGraph, CrewAI, or similar), which means owning the glue code and debugging across two systems when memory retrieval and task execution diverge.
- The project is at v0.1.0 with the open-source release flagged as new. The knowledge graph engine and temporal truth decay subsystems are advertised but lack the community-tested surface area of established memory stores. Teams building production agents that cannot tolerate evolving APIs will hit breaking changes before the interface stabilizes.
- Fact supersession logic is deterministic by design, which works cleanly for discrete facts like location or ownership — but nuanced preference evolution ("I mostly still like coffee but only in the mornings now") requires the application layer to model partial invalidation explicitly. Teams handling ambiguous or graduated state changes find themselves writing conflict-resolution logic that the engine does not provide out of the box, at which point simpler alternatives backed by relational stores start looking more tractable.
Community Reviews
Sign in to write a reviewNo reviews yet. Be the first to share your experience.
About
- Platforms
- Self-hosted binary, platform deployment
- API Available
- Yes
- Self-Hosted
- Yes
- Last Updated
- 2026-06-14T22:30:31.102Z
Best For
Who it's for
- AI agent developers needing temporal awareness
- Applications requiring fact supersession and truth decay
- Teams seeking a lightweight, self-hostable memory layer
- Multi-model workflows that must preserve state across backends
What it does well
- Maintaining evolving user preferences across chat sessions and models
- Storing and recalling structured facts such as ownership or location changes
- Providing deterministic counts and numeric answers from memory
- Distilling long chat histories into compact, queryable profiles
Integrations
Discussion Community
Sign in to commentNo discussion yet. Sign in to start the conversation.
Compare debate.tellodb
Spotted incorrect or missing data? Join our community of contributors.
Sign Up to ContributeCommunity Notes & Tips Community
Sign in to contributeBe the first to contribute. General notes, observations, gotchas, and tips from people who use this tool day-to-day.
Frequently Asked Questions
- Is debate.tellodb free?
- debate.tellodb is a paid tool. No permanent free tier is offered.
- Is debate.tellodb open source?
- No — debate.tellodb is a closed-source tool. Source code is not publicly available.
- Does debate.tellodb have an API?
- Yes. debate.tellodb exposes a developer API. See the official documentation at https://tellodb.com for details.
- Can I self-host debate.tellodb?
- Yes. debate.tellodb supports self-hosting on your own infrastructure.
- What platforms does debate.tellodb support?
- debate.tellodb is available on: Self-hosted binary, platform deployment.
Hours Saved & ROI Stories Community
Sign in to contributeBe the first to contribute. Concrete time/cost savings, with context. e.g. "Cut my code review backlog from 4h to 45m per week."
Curated lists that include this category
Standard vector stores treat every retrieved chunk as equally valid regardless of age, which breaks the moment user state evolves — addresses change, preferences flip, ownership changes. TelloDB is a memory engine that addresses this by pairing hybrid search (HNSW vector index plus BM25 full-text) with a temporal truth layer that tags older facts as stale when newer ones contradict them. The ingestion pipeline distills raw chat input into structured facts rather than storing raw conversational text, keeping the memory lattice compact and queryable. The workflow is: ingest a natural-language fact, let the distillation kernel extract structured meaning, and query against a clean profile rather than a bloated transcript store.
The differentiating feature is fact supersession combined with deterministic aggregation. When a user states they moved or bought something new, the engine resolves the conflict — marking the prior state stale and committing the new fact as active — rather than leaving contradiction resolution to the LLM at query time. The Metric Vault layer handles numeric and count-based queries (“how many cars do I own?”) through a built-in execution layer before the prompt reaches the LLM, which the vendor positions as a fix for known benchmark failures in pure vector-based memory.
TelloDB fits teams building personalization layers, stateful assistants, or multi-model pipelines where session state must survive backend switches. It ships as a single Rust binary with a self-host path and a one-click hosted deployment option, making it a candidate for air-gapped or data-sovereign deployments via its documented proxy gateway. Where it breaks: at v0.1.0, the tool is explicitly a memory substrate — it does not provide agent task planning, tool-calling scaffolding, or workflow branching. Teams that need those capabilities alongside persistent memory will be composing TelloDB with a separate agent framework and maintaining that integration themselves.