Mithril

https://docs.mithril.ai
Admin
Mithril is a self-serve cloud platform designed for machine learning research and infrastructure...
Tokens:80,541
Snippets:596
Trust Score:7
Update:3 months ago
Show doc for...
Tokens:
Raw
### Complete task configuration example

Source: https://docs.mithril.ai/mithril-cli/task-yaml.md

A comprehensive example demonstrating resource allocation, environment variables, setup, and distributed execution.

```yaml
name: llm-training

resources:
  infra: mithril
  accelerators: B200:8

config:
  mithril:
    limit_price: 32.0

num_nodes: 4

workdir: .

file_mounts:
  /data: s3://my-bucket/training-data

volumes:
  /checkpoints: my-checkpoints-volume

envs:
  WANDB_PROJECT: my-project
  WANDB_API_KEY: ${WANDB_API_KEY}

setup: |
  pip install -r requirements.txt
  wandb login

run: |
  torchrun \
    --nproc_per_node=8 \
    --nnodes=$SKYPILOT_NUM_NODES \
    train.py \
    --data /data \
    --output /checkpoints
```

--------------------------------

### GET /getting-started/account-setup.md

Source: https://docs.mithril.ai/getting-started/account-setup.md

Query the documentation dynamically by performing an HTTP GET request on the page URL with specific parameters.

```APIDOC
## GET /getting-started/account-setup.md

### Description
Perform an HTTP GET request on the current page URL to query the documentation dynamically when information is not explicitly present or clarification is needed.

### Method
GET

### Endpoint
https://docs.mithril.ai/getting-started/account-setup.md

### Parameters
#### Query Parameters
- **ask** (string) - Required - The immediate question to ask, written in natural language.
- **goal** (string) - Optional - Describes the broader end goal to help tailor the answer.
```

--------------------------------

### SSH Usage Examples

Source: https://docs.mithril.ai/mithril-cli/cli-reference/infrastructure.md

Examples for connecting to instances, running remote commands, and printing connection strings.

```bash
ml ssh my-instance                            # interactive shell
ml ssh my-instance nvidia-smi                 # run command on all nodes
ml ssh my-instance --node 0 nvidia-smi        # run on specific node
ml ssh my-instance --show                     # print ssh command
ml ssh                                        # interactive picker
```

--------------------------------

### Complete Training Configuration Example

Source: https://docs.mithril.ai/mithril-cli/task-yaml/data-and-storage.md

A comprehensive example demonstrating infrastructure selection, code synchronization, dataset mounting, and persistent volume usage.

```yaml
name: training-run

resources:
  infra: mithril/us-central5-a
  accelerators: B200:8

workdir: .

file_mounts:
  /datasets: s3://my-bucket/imagenet

volumes:
  /checkpoints: my-checkpoints-volume

run: |
  python train.py \
    --data /datasets \
    --scratch /scratch \
    --output /checkpoints
```

--------------------------------

### Kubernetes Update Examples

Source: https://docs.mithril.ai/mithril-cli/cli-reference/infrastructure.md

Usage examples for updating kubeconfig, including interactive mode and skipping backups.

```bash
ml k8s update-kubeconfig my-cluster
ml k8s update-kubeconfig my-cluster -y --no-backup
ml k8s update-kubeconfig                              # interactive picker
```

--------------------------------

### Install Kubernetes Prerequisites

Source: https://docs.mithril.ai/mithril-cli/task-yaml/infra.md

Ensure necessary tools are installed on your local machine to interact with Kubernetes clusters.

```bash
brew install kubectl socat netcat
```

--------------------------------

### ml start

Source: https://docs.mithril.ai/mithril-cli/cli-reference/workloads.md

Start one or more stopped clusters or retry provisioning for clusters in INIT state.

```APIDOC
## ml start

### Description
Start one or more stopped clusters (or retry provisioning/setup for clusters in INIT). No effect if a cluster is already UP.

### Usage
`ml start [CLUSTER ...] [OPTIONS]`

### Arguments
- **CLUSTER** (string) - Optional - Cluster name(s). Default: all clusters.

### Options
- **-a, --all** - Start all clusters.
- **-y, --yes** - Skip confirmation.
- **-i, --idle-minutes-to-autostop N** - Set autostop after N minutes of idleness.
- **--down** - Use autodown (tear down after idleness).
- **-r, --retry-until-up** - Retry until the cluster is up on availability failures.
- **-f, --force** - Start even if already UP.
```

--------------------------------

### GET /compute-and-storage/compute-quickstart.md

Source: https://docs.mithril.ai/compute-and-storage/compute-quickstart.md

Queries the documentation dynamically by asking a question related to the content.

```APIDOC
## GET /compute-and-storage/compute-quickstart.md

### Description
Perform an HTTP GET request on the documentation page URL to retrieve answers to specific questions using the documentation as a knowledge base.

### Method
GET

### Endpoint
https://docs.mithril.ai/compute-and-storage/compute-quickstart.md

### Parameters
#### Query Parameters
- **ask** (string) - Required - The immediate question to ask, written in natural language.
- **goal** (string) - Optional - Describes the broader end goal to help tailor the answer.
```

--------------------------------

### Install Mithril CLI

Source: https://docs.mithril.ai/mithril-cli/installation.md

Installs the Mithril CLI tool using the uv package manager.

```bash
# Prereq: install uv
curl -LsSf https://astral.sh/uv/install.sh | sh

# Install the Mithril CLI
uv tool install -U --refresh mithril-client
```

--------------------------------

### GET /compute-and-storage/startup-scripts.md

Source: https://docs.mithril.ai/compute-and-storage/startup-scripts.md

Queries the documentation dynamically by asking a question related to the content.

```APIDOC
## GET /compute-and-storage/startup-scripts.md

### Description
Perform an HTTP GET request on the current page URL to query the documentation dynamically.

### Method
GET

### Endpoint
https://docs.mithril.ai/compute-and-storage/startup-scripts.md

### Parameters
#### Query Parameters
- **ask** (string) - Required - The immediate question to ask.
- **goal** (string) - Optional - The broader end goal to tailor the answer.
```

--------------------------------

### Initialize Mithril CLI

Source: https://docs.mithril.ai/mithril-cli/installation.md

Commands for interactive onboarding and verifying the current setup.

```bash
ml setup # interactive onboarding
```

```bash
ml setup --check
```

--------------------------------

### Query Documentation via HTTP GET

Source: https://docs.mithril.ai/compute-and-storage/compute-quotas.md

Perform a GET request to the documentation URL with 'ask' and 'goal' parameters to retrieve context-aware information.

```http
GET https://docs.mithril.ai/compute-and-storage/compute-quotas.md?ask=<question>&goal=<endgoal>
```

--------------------------------

### Query documentation via HTTP GET

Source: https://docs.mithril.ai/access-management/porting-iam-roles-and-permissions-from-aws-or-gcp.md

Perform a GET request to the documentation URL with 'ask' and 'goal' parameters to retrieve context-aware information.

```http
GET https://docs.mithril.ai/access-management/porting-iam-roles-and-permissions-from-aws-or-gcp.md?ask=<question>&goal=<endgoal>
```

--------------------------------

### Start clusters

Source: https://docs.mithril.ai/mithril-cli/cli-reference/workloads.md

Initiate stopped clusters or retry provisioning for clusters in the INIT state.

```bash
ml start my-cluster
ml start cluster1 cluster2
ml start -a
```

--------------------------------

### Define one-time setup commands

Source: https://docs.mithril.ai/mithril-cli/task-yaml.md

Specify commands that run once when the cluster is provisioned.

```yaml
setup: |
  pip install -r requirements.txt
  apt-get update && apt-get install -y vim
```

--------------------------------

### GET /v2/instance-types

Source: https://docs.mithril.ai/compute-api/compute-api-reference/instance-types.md

Retrieves a list of all available instance types with their hardware specifications.

```APIDOC
## GET /v2/instance-types

### Description
Get all available instance types.

### Method
GET

### Endpoint
/v2/instance-types

### Response
#### Success Response (200)
- **items** (array) - A list of InstanceTypeModel objects.

#### InstanceTypeModel Schema
- **fid** (string) - Unique identifier
- **name** (string) - Name of the instance type
- **num_cpus** (integer) - Number of CPUs
- **cpu_type** (string) - Type of CPU
- **ram_gb** (integer) - RAM in GB
- **num_gpus** (integer) - Number of GPUs
- **gpu_type** (string) - Type of GPU
- **gpu_memory_gb** (integer) - GPU memory in GB
- **gpu_socket** (string) - GPU socket type
- **local_storage_gb** (integer) - Local storage in GB
- **network_type** (string, optional) - Network type
- **ib_count** (integer) - InfiniBand count
- **bridge_count** (integer) - Bridge count
```

--------------------------------

### List Kubernetes clusters

Source: https://docs.mithril.ai/mithril-cli/cli-reference/infrastructure.md

Syntax and usage examples for listing Kubernetes clusters managed by the CLI.

```bash
ml k8s list [OPTIONS]
```

```bash
ml k8s list
ml k8s list --all
ml k8s list --json
```

--------------------------------

### GET /compute-and-storage/spot-bids.md

Source: https://docs.mithril.ai/compute-and-storage/spot-bids.md

Queries the documentation for specific information by providing a question and an optional goal.

```APIDOC
## GET /compute-and-storage/spot-bids.md

### Description
Perform an HTTP GET request on the current page URL to query the documentation dynamically. This is useful when information is not explicitly present or when clarification is needed.

### Method
GET

### Endpoint
https://docs.mithril.ai/compute-and-storage/spot-bids.md

### Parameters
#### Query Parameters
- **ask** (string) - Required - The immediate question to ask, written in natural language.
- **goal** (string) - Optional - Describes the broader end goal to help tailor the answer.

### Response
#### Success Response (200)
- **answer** (string) - A direct answer to the question.
- **excerpts** (array) - Relevant excerpts from the documentation.
- **sources** (array) - Sources used to generate the answer.
```

--------------------------------

### GET /getting-started.md

Source: https://docs.mithril.ai/getting-started

Performs a dynamic query against the documentation page to retrieve answers, excerpts, and sources based on a specific question and optional goal.

```APIDOC
## GET /getting-started.md

### Description
Performs a dynamic query against the documentation page to retrieve answers, excerpts, and sources based on a specific question and optional goal.

### Method
GET

### Endpoint
https://docs.mithril.ai/getting-started.md

### Parameters
#### Query Parameters
- **ask** (string) - Required - The immediate question to ask, written in natural language.
- **goal** (string) - Optional - Describes the broader end goal to help tailor the answer.
```

--------------------------------

### View provisioning output

Source: https://docs.mithril.ai/mithril-cli/overview.md

Example output showing the CLI evaluating capacity and proposing a cluster configuration.

```text
> ml launch task.yaml
mithril-client 0.1.0
SkyPilot API server 0.1.0a4
Considered resources (2 nodes):
-------------------------------------------------------------------------------------
 INFRA                     INSTANCE   vCPUs   Mem(GB)   GPUS     COST ($)   CHOSEN
-------------------------------------------------------------------------------------
 Mithril (us-central5-a)   b200.8x    232      192       B200:8   2.25          ✔
-------------------------------------------------------------------------------------
Launching a new cluster 'sky-692d-olivier'. Proceed? [Y/n]:
• Launching... View logs: ml logs --provision sky-69a7-olivier
```

--------------------------------

### GET /v2/reservation/availability

Source: https://docs.mithril.ai/compute-api/compute-api-reference/reservations.md

Get availability information for reservations. This endpoint supports three different modes: latest_end_time (default), slots, and check.

```APIDOC
## GET /v2/reservation/availability

### Description
Get availability information for reservations. This endpoint supports three different modes for querying availability: latest_end_time, slots, and check.

### Method
GET

### Endpoint
/v2/reservation/availability

### Parameters
#### Query Parameters
- **project** (string) - Required - Project FID
- **instance_type** (string) - Required - Instance type FID
- **region** (string) - Required - Region name
- **mode** (string) - Optional - Query mode: slots, latest_end_time, or check. Defaults to latest_end_time.
- **earliest_start_time** (string) - Optional - Start of the time range to search (required for slots mode)
- **latest_end_time** (string) - Optional - End of the time range to search (required for slots mode)
- **start_time** (string) - Optional - Start time for the reservation (required for latest_end_time and check modes)
- **end_time** (string) - Optional - End time for the reservation (required for check mode)
- **quantity** (integer) - Optional - Number of instances needed (required for latest_end_time and check modes)

### Response
#### Success Response (200)
- **AvailabilitySlotModel** (array) - List of available time slots with quantities
```

--------------------------------

### GET /getting-started/welcome-to-mithril.md

Source: https://docs.mithril.ai/getting-started/welcome-to-mithril.md

Perform a dynamic query against the documentation to retrieve answers, excerpts, and sources based on a specific question and optional goal.

```APIDOC
## GET /getting-started/welcome-to-mithril.md

### Description
Queries the documentation dynamically to receive a direct answer and relevant excerpts based on the provided question.

### Method
GET

### Endpoint
https://docs.mithril.ai/getting-started/welcome-to-mithril.md

### Parameters
#### Query Parameters
- **ask** (string) - Required - The specific, self-contained question to ask the documentation.
- **goal** (string) - Optional - The broader end goal to help tailor the answer context.
```

--------------------------------

### Get Lifecycle Script Content OpenAPI Specification

Source: https://docs.mithril.ai/compute-api/compute-api-reference/lifecycle-scripts.md

Defines the GET endpoint for retrieving the content of a specific lifecycle script identified by its ID.

```json
{"openapi":"3.1.0","info":{"title":"Mithril","version":"1.0.0"},"servers":[{"url":"https://api.mithril.ai"}],"security":[{"MithrilAPIKey":[]}],"components":{"securitySchemes":{"MithrilAPIKey":{"type":"http","scheme":"bearer","bearerFormat":"fkey_<key>"}},"schemas":{"HTTPValidationError":{"properties":{"detail":{"items":{"$ref":"#/components/schemas/ValidationError"},"type":"array","title":"Detail"}},"type":"object","title":"HTTPValidationError"},"ValidationError":{"properties":{"loc":{"items":{"oneOf":[{"type":"string"},{"type":"integer"}]},"type":"array","title":"Location"},"msg":{"type":"string","title":"Message"},"type":{"type":"string","title":"Error Type"}},"type":"object","required":["loc","msg","type"],"title":"ValidationError"}}},"paths":{"/v2/lifecycle-scripts/{ls_fid}/content":{"get":{"tags":["lifecycle scripts"],"summary":"Get Lifecycle Script Content","operationId":"get_lifecycle_script_content_v2_lifecycle_scripts__ls_fid__content_get","parameters":[{"name":"ls_fid","in":"path","required":true,"schema":{"type":"string","title":"Ls Fid"}}],"responses":{"200":{"description":"Successful Response","content":{"application/json":{"schema":{}}}},"422":{"description":"Validation Error","content":{"application/json":{"schema":{"$ref":"#/components/schemas/HTTPValidationError"}}}}}}}}}
```

--------------------------------

### Get Kubernetes Clusters OpenAPI Specification

Source: https://docs.mithril.ai/compute-api/compute-api-reference/kubernetes-clusters.md

The OpenAPI definition for the Kubernetes clusters endpoint, including the cluster model schema and the GET operation.

```json
{"openapi":"3.1.0","info":{"title":"Mithril","version":"1.0.0"},"servers":[{"url":"https://api.mithril.ai"}],"security":[{"MithrilAPIKey":[]}],"components":{"securitySchemes":{"MithrilAPIKey":{"type":"http","scheme":"bearer","bearerFormat":"fkey_<key>"}},"schemas":{"KubernetesClusterModel":{"properties":{"fid":{"type":"string","title":"Fid"},"project":{"type":"string","title":"Project"},"name":{"type":"string","title":"Name"},"region":{"type":"string","title":"Region"},"created_at":{"type":"string","format":"datetime","title":"Created At"},"kube_host":{"type":"string","nullable":true},"ssh_keys":{"items":{"type":"string"},"type":"array","title":"Ssh Keys"},"instances":{"items":{"type":"string"},"type":"array","title":"Instances"},"join_command":{"type":"string","nullable":true},"status":{"type":"string","enum":["Pending","Available","Terminated"],"title":"Status"},"deleted_at":{"type":"string","format":"datetime","nullable":true},"k8s_version":{"type":"string","title":"K8S Version"},"user_fid":{"type":"string","nullable":true}},"type":"object","required":["fid","project","name","region","created_at","ssh_keys","instances","status","k8s_version"],"title":"KubernetesClusterModel"},"HTTPValidationError":{"properties":{"detail":{"items":{"$ref":"#/components/schemas/ValidationError"},"type":"array","title":"Detail"}},"type":"object","title":"HTTPValidationError"},"ValidationError":{"properties":{"loc":{"items":{"oneOf":[{"type":"string"},{"type":"integer"}]},"type":"array","title":"Location"},"msg":{"type":"string","title":"Message"},"type":{"type":"string","title":"Error Type"}},"type":"object","required":["loc","msg","type"],"title":"ValidationError"}}},"paths":{"/v2/kubernetes/clusters":{"get":{"tags":["kubernetes clusters"],"summary":"Get Kubernetes Clusters","description":"Get all Kubernetes clusters for a project","operationId":"get_kubernetes_clusters_v2_kubernetes_clusters_get","parameters":[{"name":"project","in":"query","required":true,"schema":{"type":"string","title":"Project"}}],"responses":{"200":{"description":"Successful Response","content":{"application/json":{"schema":{"type":"array","items":{"$ref":"#/components/schemas/KubernetesClusterModel"},"title":"Response Get Kubernetes Clusters V2 Kubernetes Clusters Get"}}}},"422":{"description":"Validation Error","content":{"application/json":{"schema":{"$ref":"#/components/schemas/HTTPValidationError"}}}}}}}}
```

--------------------------------

### Access startup script and logs

Source: https://docs.mithril.ai/compute-and-storage/startup-scripts.md

Commands to view the content of the initial startup script or its execution logs from within the instance.

```bash
cat /var/lib/foundry/startup_script.sh
```

```bash
cat /var/log/foundry/startup_script.log
```

--------------------------------

### GET /compute-and-storage/reserving-compute/return-credit.md

Source: https://docs.mithril.ai/compute-and-storage/reserving-compute/return-credit.md

Perform an HTTP GET request on the documentation page URL to retrieve answers to specific questions based on the documentation content.

```APIDOC
## GET /compute-and-storage/reserving-compute/return-credit.md

### Description
Perform an HTTP GET request on the current page URL with query parameters to ask questions about the documentation content.

### Method
GET

### Endpoint
https://docs.mithril.ai/compute-and-storage/reserving-compute/return-credit.md

### Parameters
#### Query Parameters
- **ask** (string) - Required - The immediate question to ask, written in natural language.
- **goal** (string) - Optional - Describes the broader end goal to help tailor the answer.
```

--------------------------------

### Launch cluster and job

Source: https://docs.mithril.ai/mithril-cli/quickstart.md

Provision a cluster and submit the initial job using the specified configuration.

```bash
ml launch task.yaml -c my-cluster
```

--------------------------------

### ml launch

Source: https://docs.mithril.ai/mithril-cli/cli-reference/workloads.md

Provisions a cluster and runs a task. If no arguments are provided, it initiates an interactive setup flow.

```APIDOC
## ml launch

### Description
Provisions a cluster and runs a task. With no arguments, it runs an interactive setup. With a task YAML or inline command, it launches the cluster and streams job logs unless --detach-run is specified.

### Synopsis
ml launch [ENTRYPOINT] [OPTIONS]

### Arguments
- **ENTRYPOINT** (string) - Optional - Path to a task YAML (.yaml/.yml) or a single bash command in quotes.

### Options
- **-c, --cluster** (string) - Optional - Cluster name. Reuses existing cluster or creates a new one.
- **--gpus** (string) - Optional - GPU type and count (e.g., A100:4).
- **--cpus** (string) - Optional - vCPU requirement (e.g., 4).
- **--memory** (string) - Optional - Memory in GB.
- **--cloud** (string) - Optional - Cloud provider.
- **--region** (string) - Optional - Region.
- **--num-nodes** (integer) - Optional - Number of nodes.
- **-i, --idle-minutes-to-autostop** (integer) - Optional - Auto-stop cluster after N minutes of idleness.
- **--down** (boolean) - Optional - Tear down the cluster after the job finishes.
- **-d, --detach-run** (boolean) - Optional - Do not stream job logs; return after job submission.
- **-r, --retry-until-up** (boolean) - Optional - Retry provisioning until the cluster is up.
- **-y, --yes** (boolean) - Optional - Skip confirmation prompts.
- **--dryrun** (boolean) - Optional - Print cluster name, task, and resources only.
- **-n, --name** (string) - Optional - Task name.
- **--workdir** (string) - Optional - Local directory to sync as the task workdir.
- **-e, --env** (string) - Optional - Set environment variables (repeatable).

### Examples
ml launch task.yaml -c mycluster
ml launch --gpus B200:8 -c dev
```

--------------------------------

### GET /v2/volumes

Source: https://docs.mithril.ai/compute-api/compute-api-reference/volumes.md

Retrieves all storage volumes for a specific project.

```APIDOC
## GET /v2/volumes

### Description
Retrieves all storage volumes for a specific project.

### Method
GET

### Endpoint
/v2/volumes

### Parameters
#### Query Parameters
- **project** (string) - Required - Identifier of the project to list volumes for.
- **region** (string) - Optional - Region filter.
- **include_deleted** (boolean) - Optional - When True, includes soft-deleted volumes in the response. Defaults to False.

### Response
#### Success Response (200)
- **Response** (array) - A list of VolumeModel objects.

#### Response Example
[
  {
    "fid": "vol-123",
    "name": "my-volume",
    "region": "us-east-1",
    "created_at": "2023-10-27T10:00:00Z",
    "deleted_at": null,
    "capacity_gb": 100,
    "project": "proj-abc",
    "interface": "Block",
    "bids": [],
    "reservations": [],
    "attachments": {}
  }
]
```

--------------------------------

### GET /v2/instances

Source: https://docs.mithril.ai/compute-api/api-overview-and-quickstart.md

Track the status of instances within a project.

```APIDOC
## GET /v2/instances

### Description
Track the status of your instances.

### Method
GET

### Endpoint
https://api.mithril.ai/v2/instances

### Query Parameters
- **project** (string) - Required - Your project FID
```

--------------------------------

### Configure persistent startup service

Source: https://docs.mithril.ai/compute-and-storage/startup-scripts.md

Example script to create a custom shell script and a systemd service that ensures execution on every boot.

```bash
#!/bin/bash

# Create the .sh file we want to run on startup
sudo tee -a /usr/local/sbin/hello-world.sh >/dev/null << 'EOF'
#!/bin/bash
echo “Hello World!”
EOF

# Make the .sh file executable
sudo chmod +x /usr/local/sbin/hello-world.sh

# Create a .service file to define a systemd service 
sudo tee -a /etc/systemd/system/startup.service >/dev/null << 'EOF'
[Unit]
Description=My Startup Script

[Service]
Type=oneshot
ExecStart=/usr/local/sbin/hello-world.sh

[Install]
WantedBy=multi-user.target
EOF

# Enable services to run on boot
sudo systemctl enable startup.service

echo "hello-world.sh will run on subsequent startups."
```

--------------------------------

### GET /v2/spot/availability

Source: https://docs.mithril.ai/compute-api/compute-api-reference/spot.md

Retrieves the current spot availability across all regions.

```APIDOC
## GET /v2/spot/availability

### Description
Get the Spot availability across all regions.

### Method
GET

### Endpoint
/v2/spot/availability

### Response
#### Success Response (200)
- **items** (array) - A list of AuctionModel objects containing availability details.

#### Response Example
[
  {
    "fid": "string",
    "instance_type": "string",
    "region": "string",
    "capacity": 0,
    "last_instance_price": "string",
    "lowest_allocated_price": null,
    "adjustable_memory": true,
    "default_image_version": null
  }
]
```

--------------------------------

### GET /v2/ssh-keys

Source: https://docs.mithril.ai/compute-api/compute-api-reference/ssh-keys.md

Retrieves all SSH keys associated with a specific project.

```APIDOC
## GET /v2/ssh-keys

### Description
Get all SSH keys for a project.

### Method
GET

### Endpoint
/v2/ssh-keys

### Parameters
#### Query Parameters
- **project** (string) - Required - The project identifier to filter SSH keys by.

### Response
#### Success Response (200)
- **Response** (array) - A list of NewSshKeyModel objects containing fid, name, project, public_key, created_at, and required status.

#### Error Response (422)
- **detail** (array) - Validation error details.
```

--------------------------------

### GET /v2/kubernetes/clusters

Source: https://docs.mithril.ai/compute-api/compute-api-reference/kubernetes-clusters.md

Retrieves a list of all Kubernetes clusters for a given project.

```APIDOC
## GET /v2/kubernetes/clusters

### Description
Get all Kubernetes clusters for a project.

### Method
GET

### Endpoint
/v2/kubernetes/clusters

### Parameters
#### Query Parameters
- **project** (string) - Required - The project identifier to filter clusters by.

### Response
#### Success Response (200)
- **Array of KubernetesClusterModel** (array) - A list of Kubernetes cluster objects.

#### Response Example
[
  {
    "fid": "string",
    "project": "string",
    "name": "string",
    "region": "string",
    "created_at": "2023-01-01T00:00:00Z",
    "ssh_keys": ["string"],
    "instances": ["string"],
    "status": "Available",
    "k8s_version": "string"
  }
]
```

--------------------------------

### Show detailed instance information

Source: https://docs.mithril.ai/mithril-cli/cli-reference/infrastructure.md

Retrieves status, configuration, and connection details for a specific bid or instance.

```bash
ml instance info [BID_NAME] [OPTIONS]
```

```bash
ml instance info my-instance
ml instance info my-instance --json
ml instance info # interactive picker
```

--------------------------------

### Inspect Kubernetes cluster details

Source: https://docs.mithril.ai/mithril-cli/cli-reference/infrastructure.md

Syntax and usage examples for retrieving detailed information about a specific Kubernetes cluster.

```bash
ml k8s info [CLUSTER] [OPTIONS]
```

```bash
ml k8s info my-cluster
ml k8s info my-cluster --json
ml k8s info                     # interactive picker
```

--------------------------------

### GET /v2/quotas

Source: https://docs.mithril.ai/compute-api/compute-api-reference/quotas.md

Retrieves all quotas for a specific project in a unified format.

```APIDOC
## GET /v2/quotas

### Description
Get all quotas for a project in unified format.

### Method
GET

### Endpoint
/v2/quotas

### Parameters
#### Query Parameters
- **project** (string) - Required - The project FID to retrieve quotas for.

### Response
#### Success Response (200)
- **QuotaModel** (array) - A list of quota objects containing fid, project, product_type, total_quantity, used_quantity, units, and name.

#### Response Example
[
  {
    "fid": "q-123",
    "project": "proj-456",
    "product_type": "Spot",
    "total_quantity": 10,
    "used_quantity": 2,
    "units": "instances",
    "name": "Spot Instances"
  }
]
```

--------------------------------

### Launch command synopsis

Source: https://docs.mithril.ai/mithril-cli/cli-reference/workloads.md

The basic syntax for the ml launch command.

```bash
ml launch [ENTRYPOINT] [OPTIONS]
```

--------------------------------

### GET /v2/me

Source: https://docs.mithril.ai/compute-api/compute-api-reference/profile.md

Retrieves the profile information for the currently authenticated user.

```APIDOC
## GET /v2/me

### Description
Get the current user's profile.

### Method
GET

### Endpoint
/v2/me

### Response
#### Success Response (200)
- **id** (string) - The user's unique identifier.
- **fid** (string) - The user's FID.
- **email** (string) - The user's email address.
- **user_name** (string, nullable) - The user's display name.
- **organization_id** (string) - The ID of the user's organization.
- **organization_role** (string, nullable) - The user's role within the organization.
```

--------------------------------

### Submit a spot bid

Source: https://docs.mithril.ai/mithril-cli/cli-reference/infrastructure.md

Use this command to initiate a spot bid for GPU instances. Requires specifying instance type, region, and maximum hourly price.

```bash
ml instance create [OPTIONS]
```

```bash
ml instance create -i b200 -r us-central5-a -m 32.0
ml instance create -i 8xa100 -r us-east1 -m 25.0 -n training -N 4
ml instance create -i b200 -r us-central5-a -m 32.0 --k8s my-cluster
ml instance create -i b200 -r us-central5-a -m 32.0 -d   # dry run
```

--------------------------------

### GET /v2/projects

Source: https://docs.mithril.ai/compute-api/compute-api-reference/projects.md

Retrieves a list of all projects that the authenticated user has access to.

```APIDOC
## GET /v2/projects

### Description
Get all projects a user has access to.

### Method
GET

### Endpoint
/v2/projects

### Response
#### Success Response (200)
- **fid** (string) - The unique identifier of the project.
- **name** (string) - The name of the project.
- **created_at** (string) - The timestamp when the project was created.

#### Response Example
[
  {
    "fid": "string",
    "name": "string",
    "created_at": "2023-01-01T00:00:00Z"
  }
]
```

--------------------------------

### Deploy and verify GPU pod

Source: https://docs.mithril.ai/compute-and-storage/mithril-compute-clusters.md

Applies the pod configuration and retrieves logs to verify GPU access.

```bash
kubectl apply -f hello-gpu.yaml
kubectl logs hello-gpu
```

--------------------------------

### GET /v2/projects

Source: https://docs.mithril.ai/compute-api/api-overview-and-quickstart.md

Retrieve a list of projects accessible to the authenticated user.

```APIDOC
## GET /v2/projects

### Description
Retrieve a list of projects you have access to.

### Method
GET

### Endpoint
https://api.mithril.ai/v2/projects
```

--------------------------------

### GET /v2/reservation/instance-types

Source: https://docs.mithril.ai/compute-api/compute-api-reference/reservations.md

Retrieves a list of instance types that are currently eligible for reservation.

```APIDOC
## GET /v2/reservation/instance-types

### Description
Get instance types available for reservation. Returns the subset of public instance types that are currently eligible for reservation.

### Method
GET

### Endpoint
/v2/reservation/instance-types

### Response
#### Success Response (200)
- **Response** (array) - A list of InstanceTypeModel objects.

#### Response Example
[
  {
    "fid": "string",
    "name": "string",
    "num_cpus": 0,
    "cpu_type": "string",
    "ram_gb": 0,
    "num_gpus": 0,
    "gpu_type": "string",
    "gpu_memory_gb": 0,
    "gpu_socket": "string",
    "local_storage_gb": 0,
    "network_type": "string",
    "ib_count": 0,
    "bridge_count": 0
  }
]
```

--------------------------------

### GET /v2/api-keys

Source: https://docs.mithril.ai/compute-api/compute-api-reference/api-keys.md

Retrieves a list of all API keys registered to the authenticated user.

```APIDOC
## GET /v2/api-keys

### Description
Get all API keys registered to the user.

### Method
GET

### Endpoint
/v2/api-keys

### Response
#### Success Response (200)
- **items** (array) - List of API key objects containing fid, name, created_at, expires_at, deactivated_at, and snippet.
```

--------------------------------

### Create Spot Bid

Source: https://docs.mithril.ai/compute-api/api-overview-and-quickstart.md

Place a bid for spot instances, including configuration for volumes, SSH keys, and startup scripts.

```python
bid_data = {
    "project": "proj_...",  # Your project FID
    "region": "us-central1-a", # Check /v2/spot/availability
    "instance_type": "it_5ECSoHQjLBzrp5YM",  # Check /v2/spot/availability
    "limit_price": "$15.50",  # Maximum price per hour you're willing to pay
    "instance_quantity": 1,
    "name": "my-training-job",
    "launch_specification": {
        "volumes": ["vol_..."],  # Your volume FID
        "ssh_keys": ["ssh_..."],  # Your SSH key FID
        "startup_script": "#!/bin/bash\necho 'Hello World'"
    }
}

response = requests.post(
    "https://api.mithril.ai/v2/spot/bids",
    headers=headers,
    json=bid_data
)
bid = response.json()
```

--------------------------------

### Create a minimal kubeconfig stub

Source: https://docs.mithril.ai/compute-and-storage/mithril-compute-clusters.md

Initialize a basic kubeconfig file if the local configuration is missing or malformed.

```bash
mkdir -p ~/.kube
cat > ~/.kube/config <<'EOF'
apiVersion: v1
kind: Config
clusters: []
contexts: []
users: []
EOF
```

--------------------------------

### Add Additional Users and Keys

Source: https://docs.mithril.ai/compute-and-storage/access-and-manage-instances.md

Commands to create a new user and configure their SSH access on the instance.

```bash
# Create a new user

sudo adduser new_user

# Switch to the new user

sudo su - new_user

# Create .ssh directory for the new user

mkdir ~/.ssh

# Change the directory permissions

chmod 700 ~/.ssh

# Create authorized_keys file

touch ~/.ssh/authorized_keys

# Add the public key to the authorized_keys file

echo "new_user_public_key" >> ~/.ssh/authorized_keys

# Change the file permissions

chmod 600 ~/.ssh/authorized_keys
```

--------------------------------

### GET /compute-and-storage/reserving-compute/pause-credit.md

Source: https://docs.mithril.ai/compute-and-storage/reserving-compute/pause-credit.md

Queries the documentation dynamically by providing a question and an optional goal.

```APIDOC
## GET /compute-and-storage/reserving-compute/pause-credit.md

### Description
Perform an HTTP GET request on the current page URL to query the documentation dynamically for information not explicitly present in the page.

### Method
GET

### Endpoint
https://docs.mithril.ai/compute-and-storage/reserving-compute/pause-credit.md

### Parameters
#### Query Parameters
- **ask** (string) - Required - The immediate question to ask, written in natural language.
- **goal** (string) - Optional - Describes the broader end goal to help tailor the answer.
```

--------------------------------

### Define environment variables

Source: https://docs.mithril.ai/mithril-cli/task-yaml.md

Set environment variables accessible during setup and execution.

```yaml
envs:
  WANDB_API_KEY: xxx
  BATCH_SIZE: 32
  DEBUG: "true"
```

--------------------------------

### Connect to Instance via SSH

Source: https://docs.mithril.ai/compute-and-storage/access-and-manage-instances.md

Use the SSH command to establish a connection to the instance using the specified key and IP address.

```bash
ssh -i /path/to/your/private/key ubuntu@<ip>
```

--------------------------------

### Query Documentation via HTTP

Source: https://docs.mithril.ai/compute-and-storage/ephemeral-storage.md

Perform an HTTP GET request to the documentation URL with 'ask' and 'goal' parameters to retrieve context-aware information.

```http
GET https://docs.mithril.ai/compute-and-storage/ephemeral-storage.md?ask=<question>&goal=<endgoal>
```

--------------------------------

### Querying documentation via HTTP

Source: https://docs.mithril.ai/compute-and-storage/compute-overview.md

Perform an HTTP GET request to the documentation URL with 'ask' and 'goal' parameters to retrieve context-aware information.

```http
GET https://docs.mithril.ai/compute-and-storage/compute-overview.md?ask=<question>&goal=<endgoal>
```

--------------------------------

### Example Preemption Signal File

Source: https://docs.mithril.ai/compute-and-storage/spot-bids/handling-spot-interruptions.md

The content of the Mithril Signal file when an instance is being preempted.

```yml
instance_status: STATUS_PREEMPTING
bid_status: Allocated
end_time: 2026-02-12 15:47:08.751581-08:00
```
Tokens:
Raw