NOISSUE - Update documentation #32
Conversation
Signed-off-by: fbugarski <[email protected]>
…tions and API docs
Signed-off-by: Filip Bugarski <[email protected]>
Signed-off-by: Filip Bugarski <[email protected]>
Signed-off-by: Filip Bugarski <[email protected]>
Signed-off-by: Filip Bugarski <[email protected]>
dborovcanin
left a comment
dborovcanin
left a comment
There was a problem hiding this comment.
I do not see a lot of documentation about proxy. Is it added in previous PRs?
docs/api/embeddings.md
Outdated
| Cube AI embeddings are generated inside **Trusted Execution Environments (TEEs)**, | ||
| ensuring that both input text and resulting vectors remain confidential. |
There was a problem hiding this comment.
Add a bit more context to it. Link to TEE, a brief explanation how it protects workload and why it's important.
There was a problem hiding this comment.
Move this to the later section, so we have a flow:
- RAG intro
- RAG explanation
- TEEs & how they help RAG and prompts
docs/api/models.md
Outdated
| Cube AI exposes language models through a **domain-scoped models registry**. | ||
| This endpoint allows clients to discover which models are available for inference | ||
| within a specific Cube AI domain. |
There was a problem hiding this comment.
What are domains, and what's domain-scoped model registry?
| - domain-scoped permissions | ||
| - per-domain model visibility | ||
|
|
||
| ### Domain Isolation |
There was a problem hiding this comment.
Again, what is domain? Workspace terminology resonates better with users in this concept, but regardless, it needs to be explained.
Signed-off-by: Filip Bugarski <[email protected]>
Signed-off-by: Filip Bugarski <[email protected]>
| ### Example Request | ||
|
|
||
| ```bash | ||
| curl -k https://localhost/proxy/<domain_id>/v1/embeddings \ |
There was a problem hiding this comment.
you can test a simple real life case and document it here
| Models in Cube AI are used by: | ||
|
|
||
| - Chat Completions | ||
| - Continue (VS Code integration) |
There was a problem hiding this comment.
we also support opencode
| ### Ollama | ||
|
|
||
| When using Ollama as a backend, models are referenced by their Ollama identifiers | ||
| (e.g. `tinyllama:1.1b`, `starcoder2:3b`). |
There was a problem hiding this comment.
superadmin can add or remove models
docs/api/models.md
Outdated
| - Model IDs are backend-specific (Ollama / vLLM) | ||
| - Models are isolated per domain | ||
| - All inference runs inside a Trusted Execution Environment (TEE) | ||
| - Models are **domain-scoped**, meaning their visibility and usage are limited |
There was a problem hiding this comment.
this is not correct models are platform wide not specific domains
Signed-off-by: Filip Bugarski <[email protected]>
Signed-off-by: Filip Bugarski <[email protected]>
Signed-off-by: Filip Bugarski <[email protected]>
There was a problem hiding this comment.
there is a guardrails doc already make changes there https://github.com/ultravioletrs/cube-docs/blob/main/docs/developer-guide/guardrails.md
No description provided.