Doso
Purpose, primary benefits, surfaces, and operating principles.
Purpose#
Doso is a place for knowledge and work that should outlast a chat. Its knowledge graph holds documents, people, projects, facts, and provenance. Its workflow graph holds definitions, skills, runs, and review. Chat is the crossing between them: a place to ask what is known and begin work, with judgment and permission applied before an agent takes the next edge.
Audience#
Builders and researchers who need durable context shared between themselves and the agents working on their behalf.
Product model#
Doso joins two systems that are often separated:
- Knowledge preserves sources, facts, relationships, and provenance.
- Workflows preserve repeatable methods, capabilities, runs, and review.
Chat connects the two. It can use existing knowledge, begin new work, and return the result with its evidence and execution history intact.
The instance is the API and its database. Web and Desktop are surfaces that talk to that instance.
Surfaces#
| Surface | Job |
|---|---|
| Chat | Ask with evidence, remember useful context, and start graph-backed work |
| Knowledge | Explore Memories, Library sources, entities, relationships, and provenance |
| Workflows | Build reusable definitions and inspect each run |
| Settings | Configure appearance, models, ontology, connections, identity, and the instance URL |
Primary benefits#
- Your graphs stay yours. The knowledge and workflow graphs you build over time remain yours, always. No vendor lock-in.
- Your connections. Reach hosted tools through your own Composio account, or pair a desktop runtime to CLIs that stay on the host.
- Host anywhere. The same web → API → Postgres stack runs locally or in any cloud, in any country.
- Any inference. Choose any provider and any model. Swap when you want.
Product principles#
- One instance, two surfaces. API + database are the core. Web and Desktop do not keep a second product database.
- Authorization is always on. A single-user compose instance still signs in and evaluates policy.
- Evidence before fluency. Answers retain citations and provenance.
- Judgment at the edges. Consequential actions pause for the right approval.
- Definitions outlast runs. A workflow is reusable structure; a run is one inspectable attempt.
- General capabilities, specific workflows. The engine stays reusable while each workflow carries the recipe for its own job.