Go: implement Embed (embeddings) in Mistral driver (#14807)
### What problem does this PR solve?
The Mistral Go driver landed in #14805 with chat, list models, and check
connection. `Embed` was left as a stub that returns `"not implemented"`.
This PR fills the gap.
`conf/models/mistral.json` did not list any embedding model out of the
box, so a tenant who wanted to use Mistral end to end (chat +
embeddings) could not run an embedding call. This PR adds
`mistral-embed` to the config and a real `/v1/embeddings`
implementation.
### What this PR includes
- `conf/models/mistral.json`: add `"embedding": "embeddings"` under
`url_suffix` so the driver can build the URL from config (matches the
`URLSuffix.Embedding` field already used by openai, siliconflow,
zhipu-ai), and add a `mistral-embed` entry under `models`
(1024-dimensional vectors, 8192 max input tokens).
- `internal/entity/models/mistral.go`: replace the `Embed` stub with a
real implementation that POSTs to `/v1/embeddings`. Adds local response
types `mistralEmbeddingData` and `mistralEmbeddingResponse`.
No factory change. No interface change.
### How the implementation works
- Validate `apiConfig`, the API key, and the model name. Use the
existing `baseURLForRegion` helper so an unknown region fails fast with
a clear error.
- Wrap the request with `context.WithTimeout(nonStreamCallTimeout)` so
the call has a clear deadline. Same pattern as `ChatWithMessages` and
`ListModels` already use in this file.
- Send all input texts in one request. The Mistral API accepts the
`input` field as an array.
- Parse `data[*].embedding` and copy each slice into a `[]EmbeddingData`
indexed by `data[*].index` so the output order matches the input order
even if the API returns items in a different order.
- An empty input slice returns `[]EmbeddingData{}` with no HTTP call.
- Non-200 responses propagate the upstream status line and body.
- A final pass checks that every input slot got a vector. If any slot is
still empty, return a clear error so the caller does not silently use a
zero vector.
### Note on stacking
This PR builds on #14805 (the Mistral driver). Until #14805 merges, this
PR's diff on GitHub will include both that PR's commits and this one.
After #14805 lands on `main`, GitHub will auto-reduce this PR to only
the `Embed` changes (one commit, ~111 line diff in `mistral.go` plus 8
lines in `mistral.json`).
### Type of change
- [x] New Feature (non-breaking change which adds functionality)
### How was this tested?
- `go build ./internal/entity/models/...` returns exit 0 on go 1.25 (the
`go.mod` minimum).
- The full method set on `MistralModel` still matches the `ModelDriver`
interface.
- Pattern parity with the existing OpenAI Embed implementation
(`internal/entity/models/openai.go`).
Closes #14806
Depends on #14805
Tracking: #14736
---------
Co-authored-by: Jin Hai <haijin.chn@gmail.com>
2026-05-11 23:45:48 -10:00
|
|
|
{
|
|
|
|
|
"name": "Mistral",
|
|
|
|
|
"url": {
|
2026-05-15 18:41:43 +08:00
|
|
|
"default": "https://api.mistral.ai"
|
Go: implement Embed (embeddings) in Mistral driver (#14807)
### What problem does this PR solve?
The Mistral Go driver landed in #14805 with chat, list models, and check
connection. `Embed` was left as a stub that returns `"not implemented"`.
This PR fills the gap.
`conf/models/mistral.json` did not list any embedding model out of the
box, so a tenant who wanted to use Mistral end to end (chat +
embeddings) could not run an embedding call. This PR adds
`mistral-embed` to the config and a real `/v1/embeddings`
implementation.
### What this PR includes
- `conf/models/mistral.json`: add `"embedding": "embeddings"` under
`url_suffix` so the driver can build the URL from config (matches the
`URLSuffix.Embedding` field already used by openai, siliconflow,
zhipu-ai), and add a `mistral-embed` entry under `models`
(1024-dimensional vectors, 8192 max input tokens).
- `internal/entity/models/mistral.go`: replace the `Embed` stub with a
real implementation that POSTs to `/v1/embeddings`. Adds local response
types `mistralEmbeddingData` and `mistralEmbeddingResponse`.
No factory change. No interface change.
### How the implementation works
- Validate `apiConfig`, the API key, and the model name. Use the
existing `baseURLForRegion` helper so an unknown region fails fast with
a clear error.
- Wrap the request with `context.WithTimeout(nonStreamCallTimeout)` so
the call has a clear deadline. Same pattern as `ChatWithMessages` and
`ListModels` already use in this file.
- Send all input texts in one request. The Mistral API accepts the
`input` field as an array.
- Parse `data[*].embedding` and copy each slice into a `[]EmbeddingData`
indexed by `data[*].index` so the output order matches the input order
even if the API returns items in a different order.
- An empty input slice returns `[]EmbeddingData{}` with no HTTP call.
- Non-200 responses propagate the upstream status line and body.
- A final pass checks that every input slot got a vector. If any slot is
still empty, return a clear error so the caller does not silently use a
zero vector.
### Note on stacking
This PR builds on #14805 (the Mistral driver). Until #14805 merges, this
PR's diff on GitHub will include both that PR's commits and this one.
After #14805 lands on `main`, GitHub will auto-reduce this PR to only
the `Embed` changes (one commit, ~111 line diff in `mistral.go` plus 8
lines in `mistral.json`).
### Type of change
- [x] New Feature (non-breaking change which adds functionality)
### How was this tested?
- `go build ./internal/entity/models/...` returns exit 0 on go 1.25 (the
`go.mod` minimum).
- The full method set on `MistralModel` still matches the `ModelDriver`
interface.
- Pattern parity with the existing OpenAI Embed implementation
(`internal/entity/models/openai.go`).
Closes #14806
Depends on #14805
Tracking: #14736
---------
Co-authored-by: Jin Hai <haijin.chn@gmail.com>
2026-05-11 23:45:48 -10:00
|
|
|
},
|
|
|
|
|
"url_suffix": {
|
2026-05-15 18:41:43 +08:00
|
|
|
"chat": "v1/chat/completions",
|
|
|
|
|
"models": "v1/models",
|
|
|
|
|
"embedding": "v1/embeddings",
|
|
|
|
|
"ocr": "v1/ocr"
|
Go: implement Embed (embeddings) in Mistral driver (#14807)
### What problem does this PR solve?
The Mistral Go driver landed in #14805 with chat, list models, and check
connection. `Embed` was left as a stub that returns `"not implemented"`.
This PR fills the gap.
`conf/models/mistral.json` did not list any embedding model out of the
box, so a tenant who wanted to use Mistral end to end (chat +
embeddings) could not run an embedding call. This PR adds
`mistral-embed` to the config and a real `/v1/embeddings`
implementation.
### What this PR includes
- `conf/models/mistral.json`: add `"embedding": "embeddings"` under
`url_suffix` so the driver can build the URL from config (matches the
`URLSuffix.Embedding` field already used by openai, siliconflow,
zhipu-ai), and add a `mistral-embed` entry under `models`
(1024-dimensional vectors, 8192 max input tokens).
- `internal/entity/models/mistral.go`: replace the `Embed` stub with a
real implementation that POSTs to `/v1/embeddings`. Adds local response
types `mistralEmbeddingData` and `mistralEmbeddingResponse`.
No factory change. No interface change.
### How the implementation works
- Validate `apiConfig`, the API key, and the model name. Use the
existing `baseURLForRegion` helper so an unknown region fails fast with
a clear error.
- Wrap the request with `context.WithTimeout(nonStreamCallTimeout)` so
the call has a clear deadline. Same pattern as `ChatWithMessages` and
`ListModels` already use in this file.
- Send all input texts in one request. The Mistral API accepts the
`input` field as an array.
- Parse `data[*].embedding` and copy each slice into a `[]EmbeddingData`
indexed by `data[*].index` so the output order matches the input order
even if the API returns items in a different order.
- An empty input slice returns `[]EmbeddingData{}` with no HTTP call.
- Non-200 responses propagate the upstream status line and body.
- A final pass checks that every input slot got a vector. If any slot is
still empty, return a clear error so the caller does not silently use a
zero vector.
### Note on stacking
This PR builds on #14805 (the Mistral driver). Until #14805 merges, this
PR's diff on GitHub will include both that PR's commits and this one.
After #14805 lands on `main`, GitHub will auto-reduce this PR to only
the `Embed` changes (one commit, ~111 line diff in `mistral.go` plus 8
lines in `mistral.json`).
### Type of change
- [x] New Feature (non-breaking change which adds functionality)
### How was this tested?
- `go build ./internal/entity/models/...` returns exit 0 on go 1.25 (the
`go.mod` minimum).
- The full method set on `MistralModel` still matches the `ModelDriver`
interface.
- Pattern parity with the existing OpenAI Embed implementation
(`internal/entity/models/openai.go`).
Closes #14806
Depends on #14805
Tracking: #14736
---------
Co-authored-by: Jin Hai <haijin.chn@gmail.com>
2026-05-11 23:45:48 -10:00
|
|
|
},
|
|
|
|
|
"class": "mistral",
|
|
|
|
|
"models": [
|
|
|
|
|
{
|
|
|
|
|
"name": "mistral-large-latest",
|
|
|
|
|
"max_tokens": 128000,
|
|
|
|
|
"model_types": [
|
|
|
|
|
"chat"
|
|
|
|
|
]
|
|
|
|
|
},
|
|
|
|
|
{
|
|
|
|
|
"name": "mistral-medium-latest",
|
|
|
|
|
"max_tokens": 128000,
|
|
|
|
|
"model_types": [
|
|
|
|
|
"chat"
|
|
|
|
|
]
|
|
|
|
|
},
|
|
|
|
|
{
|
|
|
|
|
"name": "mistral-small-latest",
|
|
|
|
|
"max_tokens": 128000,
|
|
|
|
|
"model_types": [
|
|
|
|
|
"chat"
|
|
|
|
|
]
|
|
|
|
|
},
|
|
|
|
|
{
|
|
|
|
|
"name": "ministral-8b-latest",
|
|
|
|
|
"max_tokens": 128000,
|
|
|
|
|
"model_types": [
|
|
|
|
|
"chat"
|
|
|
|
|
]
|
|
|
|
|
},
|
|
|
|
|
{
|
|
|
|
|
"name": "ministral-3b-latest",
|
|
|
|
|
"max_tokens": 128000,
|
|
|
|
|
"model_types": [
|
|
|
|
|
"chat"
|
|
|
|
|
]
|
|
|
|
|
},
|
|
|
|
|
{
|
|
|
|
|
"name": "pixtral-large-latest",
|
|
|
|
|
"max_tokens": 128000,
|
|
|
|
|
"model_types": [
|
|
|
|
|
"chat",
|
|
|
|
|
"vision"
|
|
|
|
|
]
|
|
|
|
|
},
|
|
|
|
|
{
|
|
|
|
|
"name": "codestral-latest",
|
|
|
|
|
"max_tokens": 256000,
|
|
|
|
|
"model_types": [
|
|
|
|
|
"chat"
|
|
|
|
|
]
|
|
|
|
|
},
|
|
|
|
|
{
|
|
|
|
|
"name": "open-mistral-nemo",
|
|
|
|
|
"max_tokens": 128000,
|
|
|
|
|
"model_types": [
|
|
|
|
|
"chat"
|
|
|
|
|
]
|
|
|
|
|
},
|
|
|
|
|
{
|
|
|
|
|
"name": "open-mistral-7b",
|
|
|
|
|
"max_tokens": 32000,
|
|
|
|
|
"model_types": [
|
|
|
|
|
"chat"
|
|
|
|
|
]
|
|
|
|
|
},
|
|
|
|
|
{
|
|
|
|
|
"name": "open-mixtral-8x7b",
|
|
|
|
|
"max_tokens": 32000,
|
|
|
|
|
"model_types": [
|
|
|
|
|
"chat"
|
|
|
|
|
]
|
|
|
|
|
},
|
|
|
|
|
{
|
|
|
|
|
"name": "open-mixtral-8x22b",
|
|
|
|
|
"max_tokens": 64000,
|
|
|
|
|
"model_types": [
|
|
|
|
|
"chat"
|
|
|
|
|
]
|
|
|
|
|
},
|
|
|
|
|
{
|
|
|
|
|
"name": "mistral-embed",
|
|
|
|
|
"max_tokens": 8192,
|
|
|
|
|
"model_types": [
|
|
|
|
|
"embedding"
|
|
|
|
|
]
|
2026-05-15 18:41:43 +08:00
|
|
|
},
|
|
|
|
|
{
|
|
|
|
|
"name": "mistral-ocr-2512",
|
|
|
|
|
"max_tokens": 8192,
|
|
|
|
|
"model_types": [
|
|
|
|
|
"ocr"
|
|
|
|
|
]
|
Go: implement Embed (embeddings) in Mistral driver (#14807)
### What problem does this PR solve?
The Mistral Go driver landed in #14805 with chat, list models, and check
connection. `Embed` was left as a stub that returns `"not implemented"`.
This PR fills the gap.
`conf/models/mistral.json` did not list any embedding model out of the
box, so a tenant who wanted to use Mistral end to end (chat +
embeddings) could not run an embedding call. This PR adds
`mistral-embed` to the config and a real `/v1/embeddings`
implementation.
### What this PR includes
- `conf/models/mistral.json`: add `"embedding": "embeddings"` under
`url_suffix` so the driver can build the URL from config (matches the
`URLSuffix.Embedding` field already used by openai, siliconflow,
zhipu-ai), and add a `mistral-embed` entry under `models`
(1024-dimensional vectors, 8192 max input tokens).
- `internal/entity/models/mistral.go`: replace the `Embed` stub with a
real implementation that POSTs to `/v1/embeddings`. Adds local response
types `mistralEmbeddingData` and `mistralEmbeddingResponse`.
No factory change. No interface change.
### How the implementation works
- Validate `apiConfig`, the API key, and the model name. Use the
existing `baseURLForRegion` helper so an unknown region fails fast with
a clear error.
- Wrap the request with `context.WithTimeout(nonStreamCallTimeout)` so
the call has a clear deadline. Same pattern as `ChatWithMessages` and
`ListModels` already use in this file.
- Send all input texts in one request. The Mistral API accepts the
`input` field as an array.
- Parse `data[*].embedding` and copy each slice into a `[]EmbeddingData`
indexed by `data[*].index` so the output order matches the input order
even if the API returns items in a different order.
- An empty input slice returns `[]EmbeddingData{}` with no HTTP call.
- Non-200 responses propagate the upstream status line and body.
- A final pass checks that every input slot got a vector. If any slot is
still empty, return a clear error so the caller does not silently use a
zero vector.
### Note on stacking
This PR builds on #14805 (the Mistral driver). Until #14805 merges, this
PR's diff on GitHub will include both that PR's commits and this one.
After #14805 lands on `main`, GitHub will auto-reduce this PR to only
the `Embed` changes (one commit, ~111 line diff in `mistral.go` plus 8
lines in `mistral.json`).
### Type of change
- [x] New Feature (non-breaking change which adds functionality)
### How was this tested?
- `go build ./internal/entity/models/...` returns exit 0 on go 1.25 (the
`go.mod` minimum).
- The full method set on `MistralModel` still matches the `ModelDriver`
interface.
- Pattern parity with the existing OpenAI Embed implementation
(`internal/entity/models/openai.go`).
Closes #14806
Depends on #14805
Tracking: #14736
---------
Co-authored-by: Jin Hai <haijin.chn@gmail.com>
2026-05-11 23:45:48 -10:00
|
|
|
}
|
|
|
|
|
]
|
|
|
|
|
}
|