diff --git a/docs/README.md b/docs/README.md new file mode 100644 index 00000000..84f7dd8a --- /dev/null +++ b/docs/README.md @@ -0,0 +1,5 @@ +# Documentation + +- [Modelfile](./modelfile.md) +- [How to develop Ollama](./development.md) +- [API](./api.md) diff --git a/docs/api.md b/docs/api.md new file mode 100644 index 00000000..27eb4031 --- /dev/null +++ b/docs/api.md @@ -0,0 +1,376 @@ +[Documentation Home](./README.md) + +# API + +- [Generate a Prompt](#generate-a-prompt) +- [Create a Model](#create-a-model) +- [List Local Models](#list-local-models) +- [Copy a Model](#copy-a-model) +- [Delete a Model](#delete-a-model) +- [Pull a Model](#pull-a-model) + +## Things to keep in mind when using the API + +### Model name format + +The model name format today is model:tag`. Some examples are `orca:3b-q4_1` and `llama2:70b`. The tag is optional and if not provided will default to `latest`. The tag is used to identify a specific version. + +### Durations + +All durations are in nanoseconds. + +## Generate a Prompt + +**POST /api/generate** + +### Description + +**Generate** is the main endpoint that you will use when working with Ollama. This is used to generate a response to a prompt sent to a model. This is a streaming endpoint, so will be a series of responses. The final response will include the context and what is usually seen in the output from verbose mode. + +### Request + +The **Generate** endpoint takes a JSON object with the following fields: + +```JSON +{ + "model": "site/namespace/model:tag", + "prompt": "You are a software engineer working on building docs for Ollama.", + "options": { + "temperature": 0.7, + } +} +``` + +**Options** can include any of the parameters listed in the [Modelfile](./modelfile.mdvalid-parameters-and-values) documentation. The only required parameter is **model**. If no **prompt** is provided, the model will generate a response to an empty prompt. If no **options** are provided, the model will use the default options from the Modelfile of the parent model. + +### Response + +The response is a stream of JSON objects with the following fields: + +```JSON +{ + "model": "modelname", + "created_at": "2023-08-04T08:52:19.385406455-07:00" + "response": "the current token", + "done": false +} +``` + +The final response in the stream also includes the context and what is usually seen in the output from verbose mode. For example: + +```JSON +{ + "model":"orca", + "created_at":"2023-08-04T19:22:45.499127Z", + "done":true, + "total_duration":5589157167, + "load_duration":3013701500, + "sample_count":114, + "sample_duration":81442000, + "prompt_eval_count":46, + "prompt_eval_duration":1160282000, + "eval_count":113, + "eval_duration":1325948000 +} +``` + +| field | description | +| -------------------- | ------------------------------------------------------- | +| model | the name of the model | +| created_at | the time the response was generated | +| response | the current token | +| done | whether the response is complete | +| total_duration | total time in nanoseconds spent generating the response | +| load_duration | time spent in nanoseconds loading the model | +| sample_count | number of samples generated | +| sample_duration | time spent generating samples | +| prompt_eval_count | number of times the prompt was evaluated | +| prompt_eval_duration | time spent in nanoseconds evaluating the prompt | +| eval_count | number of times the response was evaluated | +| eval_duration | time in nanoseconds spent evaluating the response | + +### Example + +#### Request + +```shell +curl -X POST 'http://localhost:11434/api/generate' -d \ +'{ + "model": "orca", + "prompt": "why is the sky blue" +}' +``` + +#### Response + +```json +{"model":"orca","created_at":"2023-08-04T19:22:44.085127Z","response":" The","done":false} +{"model":"orca","created_at":"2023-08-04T19:22:44.176425Z","response":" sky","done":false} +{"model":"orca","created_at":"2023-08-04T19:22:44.18883Z","response":" appears","done":false} +{"model":"orca","created_at":"2023-08-04T19:22:44.200852Z","response":" blue","done":false} +{"model":"orca","created_at":"2023-08-04T19:22:44.213644Z","response":" because","done":false} +{"model":"orca","created_at":"2023-08-04T19:22:44.225706Z","response":" of","done":false} +{"model":"orca","created_at":"2023-08-04T19:22:44.237686Z","response":" a","done":false} +... +{"model":"orca","created_at":"2023-08-04T19:22:45.487113Z","response":".","done":false} +{"model":"orca","created_at":"2023-08-04T19:22:45.499127Z","done":true,"total_duration":5589157167,"load_duration":3013701500,"sample_count":114,"sample_duration":81442000,"prompt_eval_count":46,"prompt_eval_duration":1160282000,"eval_count":113,"eval_duration":1325948000} +``` + +## Create a Model + +**POST /api/create** + +### Description + +**Create** takes a path to a Modelfile and creates a model. The Modelfile is documented [here](./modelfile.md). + +### Request + +The **Create** endpoint takes a JSON object with the following fields: + +```JSON +{ + "name": "modelname", + "path": "absolute path to Modelfile" +} +``` + +### Response + +The response is a stream of JSON objects that have a single key/value pair for status. For example: + +```JSON +{ + "status": "parsing modelfile" +} +``` + +### Example + +#### Request + +```shell +curl --location --request POST 'http://localhost:11434/api/create' \ +--header 'Content-Type: text/plain' \ +--data-raw '{ + "name": "myCoolModel", + "path": "/Users/matt/ollamamodelfiles/sentiments" +}' +``` + +#### Response + +```JSON +{"status":"parsing modelfile"} +{"status":"looking for model"} +{"status":"creating model template layer"} +{"status":"creating config layer"} +{"status":"using already created layer sha256:e84705205f71dd55be7b24a778f248f0eda9999a125d313358c087e092d83148"} +{"status":"using already created layer sha256:93ca9b3d83dc541f11062c0b994ae66a7b327146f59a9564aafef4a4c15d1ef5"} +{"status":"writing layer sha256:d3fe6fb39620a477da7720c5fa00abe269a018a9675a726320e18122b7142ee7"} +{"status":"writing layer sha256:16cc83359b0395026878b41662f7caef433f5260b5d49a3257312b6417b7d8a8"} +{"status":"writing manifest"} +{"status":"success"} +``` + +## List Local Models + +**GET /api/tags** + +### Description + +**List** will list out the models that are available locally. + +### Request + +The **List** endpoint takes no parameters and is a simple GET request. + +### Response + +The response is a JSON object with a single key/value pair for models. For example: + +```JSON +{ + "models": [ + { + "name": "modelname:tags", + "modified_at": "2023-08-04T08:52:19.385406455-07:00", + "size": 7323310500 + } + ] +} +``` + +### Example + +#### Request + +```shell +curl 'http://localhost:11434/api/tags' +``` + +#### Response + +```JSON +{ + "models": [ + { + "name": "llama2:70b", + "modified_at": "2023-08-04T08:52:19.385406455-07:00", + "size": 38871966966 + }, + { + "name": "llama2:70b-chat-q4_0", + "modified_at": "2023-08-04T09:21:27.703371485-07:00", + "size": 38871974480 + }, + { + "name": "midjourney-prompter:latest", + "modified_at": "2023-08-04T08:45:46.399609053-07:00", + "size": 7323311708 + }, + { + "name": "raycast_orca:3b", + "modified_at": "2023-08-04T06:23:20.10832636-07:00", + "size": 1928446602 + }, + { + "name": "stablebeluga:13b-q4_K_M", + "modified_at": "2023-08-04T09:48:26.416547463-07:00", + "size": 7865679045 + } + ] +} +``` + +## Copy a Model + +**POST /api/copy** + +### Description + +**Copy** will copy a model from one name to another. This is useful for creating a new model from an existing model. It is often used as the first step to renaming a model. + +### Request + +The **Copy** endpoint takes a JSON object with the following fields: + +```JSON +{ + "source": "modelname", + "destination": "newmodelname" +} +``` + +### Response + +There is no response other than a 200 status code. + +### Example + +#### Request + +```shell +curl -X POST 'http://localhost:11434/api/copy' -d \ +'{ + "source": "MyCoolModel", + "destination": "ADifferentModel" +}' +``` + +#### Response + +No response is returned other than a 200 status code. + +## Delete a Model + +**DELETE /api/delete** + +### Description + +**Delete** will delete a model from the local machine. This is useful for cleaning up models that are no longer needed. + +### Request + +The **Delete** endpoint takes a JSON object with a single key/value pair for modelname. For example: + +```JSON +{ + "model": "modelname" +} +``` + +### Response + +No response is returned other than a 200 status code. + +### Example + +#### Request + +```shell +curl -X DELETE 'http://localhost:11434/api/delete' -d \ +'{ + "name": "adifferentModel" +}' +``` + +#### Response + +No response is returned other than a 200 status code. + +## Pull a Model + +**POST /api/pull** + +### Description + +**Pull** will pull a model from a remote registry. This is useful for getting a model from the Ollama registry and in the future from alternate registries. + +### Request + +The **Pull** endpoint takes a JSON object with the following fields: + +```JSON +{ + "name": "modelname" +} +``` + +### Response + +The response is a stream of JSON objects with the following format: + +```JSON +{ + "status":"downloading digestname", + "digest":"digestname", + "total":2142590208 +} +``` + +### Example + +#### Request + +```shell +curl -X POST 'http://localhost:11434/api/pull' -d \ +'{ + "name": "orca:3b-q4_1" +}' +``` + +#### Response + +```JSON +{"status":"pulling manifest"} +{"status":"downloading sha256:63151c63f792939bb4a40b35f37ea06e047c02486399d1742113aaefd0d33e29","digest":"sha256:63151c63f792939bb4a40b35f37ea06e047c02486399d1742113aaefd0d33e29","total":2142590208} +{"status":"downloading sha256:63151c63f792939bb4a40b35f37ea06e047c02486399d1742113aaefd0d33e29","digest":"sha256:63151c63f792939bb4a40b35f37ea06e047c02486399d1742113aaefd0d33e29","total":2142590208,"completed":1048576} +... +{"status":"downloading sha256:20714f2ebe4be44313358bfa58556d783652398ed47f12178914c706c4ad12c4","digest":"sha256:20714f2ebe4be44313358bfa58556d783652398ed47f12178914c706c4ad12c4","total":299} +{"status":"downloading sha256:20714f2ebe4be44313358bfa58556d783652398ed47f12178914c706c4ad12c4","digest":"sha256:20714f2ebe4be44313358bfa58556d783652398ed47f12178914c706c4ad12c4","total":299,"completed":299} +{"status":"verifying sha256 digest"} +{"status":"writing manifest"} +{"status":"success"} + +```