Pioneer

https://docs.pioneer.ai
Admin
Pioneer is an AI fine-tuning and inference platform that lets you fine-tune open-source LLMs and NER...
Tokens:71,841
Snippets:729
Trust Score:6.1
Update:2 months ago
Show doc for...
Tokens:
Raw
### Install and Verify Bun

Source: https://docs.pioneer.ai/CLI-Installation

Install the Bun runtime and verify the installation by checking the version.

```bash
curl -fsSL https://bun.sh/install | bash
```

```bash
bun --version
```

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

### Install Mintlify skill

Source: https://docs.pioneer.ai/AGENTS

Use this command to install the Mintlify skill for product knowledge and configuration support.

```bash
npx skills add https://mintlify.com/docs
```

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

### Run Hermes Agent Setup Command

Source: https://docs.pioneer.ai/hermes

Executes a script to fetch the Pioneer model catalog, filter aliases, and update the Hermes configuration file. Requires jq and ruby installed on the system.

```bash
: "${PIONEER_API_KEY:?Set PIONEER_API_KEY first}"

command -v jq >/dev/null || { echo "jq is required"; exit 1; }
command -v ruby >/dev/null || { echo "ruby is required"; exit 1; }

CONFIG_FILE="$(hermes config path)"
CONFIG_TMP="$(mktemp)"
MODELS_JSON="$(mktemp)"
trap 'rm -f "$CONFIG_TMP" "$MODELS_JSON"' EXIT

mkdir -p "$(dirname "$CONFIG_FILE")"
[ -f "$CONFIG_FILE" ] || printf '{}\n' > "$CONFIG_FILE"
cp "$CONFIG_FILE" "$CONFIG_FILE.bak.$(date +%Y%m%d%H%M%S)"

hermes config set PIONEER_API_KEY "$PIONEER_API_KEY"

curl -fsS "https://api.pioneer.ai/v1/models" \
  -H "Authorization: Bearer $PIONEER_API_KEY" \
  | jq '
    def dedupe:
      reduce .[] as $item ([]; if index($item) then . else . + [$item] end);

    def catalog_models:
      (.models // []) as $models
      | if (($models | type) == "array" and ($models | length) > 0) then $models else (.data // []) end;

    def model_id:
      .slug // .id;

    [
      catalog_models[]
      | select(.deprecated != true)
      | model_id
      | select(type == "string" and length > 0)
      # Hide Claude Code discovery aliases so Hermes does not show every model twice.
      | select(startswith("anthropic/") | not)
    ] | dedupe
  ' > "$MODELS_JSON"

ruby -ryaml -rjson -e '
  config_path, models_path, out_path = ARGV
  cfg = File.exist?(config_path) ? (YAML.safe_load(File.read(config_path), aliases: true) || {}) : {}
  abort "#{config_path} must contain a YAML mapping" unless cfg.is_a?(Hash)

  models = JSON.parse(File.read(models_path))
  cfg["providers"] = {} unless cfg["providers"].is_a?(Hash)
  cfg["providers"]["pioneer"] = {
    "name" => "Pioneer",
    "base_url" => "https://api.pioneer.ai/v1",
    "key_env" => "PIONEER_API_KEY",
    "api_mode" => "chat_completions",
    "discover_models" => false,
    "default_model" => "claude-opus-5",
    "models" => models
  }

  cfg["model"] = {} unless cfg["model"].is_a?(Hash)
  cfg["model"]["provider"] = "pioneer"
  cfg["model"]["default"] = "claude-opus-5"
  cfg["model"]["base_url"] = "https://api.pioneer.ai/v1"
  cfg["model"]["api_mode"] = "chat_completions"

  File.write(out_path, YAML.dump(cfg))
' "$CONFIG_FILE" "$MODELS_JSON" "$CONFIG_TMP"

mv "$CONFIG_TMP" "$CONFIG_FILE"
chmod 600 "$CONFIG_FILE"
```

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

### Install Pioneer CLI

Source: https://docs.pioneer.ai/CLI-Installation

Install the Pioneer CLI globally using npm.

```bash
npm install -g @fastino-ai/pioneer-cli
```

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

### Start Hermes Agent

Source: https://docs.pioneer.ai/hermes

Launch the Hermes Agent after configuration.

```bash
hermes
```

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

### Start a generation job

Source: https://docs.pioneer.ai/api-reference/synthetic-data

Starts an asynchronous job that generates labeled training examples and stores them in a named dataset. Returns a job ID for polling.

```APIDOC
## POST /generate

### Description
Starts an asynchronous job that generates labeled training examples and stores them in a named dataset. Returns a job ID you can use to poll for completion.

### Method
POST

### Endpoint
/generate

### Parameters
#### Request Body
- **task_type** (string) - Required - The type of task to generate data for. Accepted values: `ner`, `classification`, `decoder`.
- **dataset_name** (string) - Required - The name of the dataset to create or append to. If a dataset with this name already exists, new examples are added as a new version.
- **num_examples** (number) - Optional - Number of labeled examples to generate.
- **labels** (string[]) - Optional - List of label strings for NER or classification tasks. For NER, these are entity type names (e.g. `"person"`, `"organization"`). For classification, these are the class names.
- **domain_description** (string) - Optional - A natural-language description of the domain or topic for the generated examples. Providing a detailed description improves example quality and relevance.
- **classified_examples** (object[]) - Optional - Few-shot examples with labels to guide generation for classification tasks.
- **prompt** (string) - Optional - Custom instruction prompt to control generation style for decoder tasks.

### Request Example
```bash
curl -X POST https://api.pioneer.ai/generate \
  -H "X-API-Key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d 
    {
      "task_type": "ner",
      "dataset_name": "my-ner-dataset",
      "labels": ["person", "company", "product"],
      "num_examples": 100,
      "domain_description": "Tech industry news articles"
    }
```

### Response
#### Success Response (200)
- **job_id** (string) - Unique identifier for the generation job. Use this with `GET /generate/jobs/:job_id` to poll for status.
- **status** (string) - Initial job status, typically `queued`.
```

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

### Verify CLI Installation

Source: https://docs.pioneer.ai/CLI-Installation

Check the CLI version and display help information.

```bash
pioneer --version
pioneer --help
```

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

### Install OpenClaw

Source: https://docs.pioneer.ai/openclaw

Install the latest version of OpenClaw globally via npm.

```shellscript
npm install -g openclaw@latest
```

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

### Install Routing-Savings Hook

Source: https://docs.pioneer.ai/codex

Script to configure environment variables and install the Python hook for tracking Pioneer session savings.

```bash
: "${PIONEER_API_KEY:?Set PIONEER_API_KEY first}"
mkdir -p ~/.pioneer/hooks ~/.pioneer/state

# Credentials the hook reads (sourcing is optional - the hook also reads this file directly)
cat > ~/.pioneer/codex-env <<EOF
export PIONEER_API_KEY="$PIONEER_API_KEY"
export PIONEER_BASE_URL="https://api.pioneer.ai/v1"
EOF
chmod 600 ~/.pioneer/codex-env

# The Codex Stop hook
cat > ~/.pioneer/hooks/show-pioneer-signals.py <<'PIONEER_CODEX_HOOK'
#!/usr/bin/env python3
"""Codex Stop hook that surfaces how much pioneer/auto routing saved.

Codex talks to Pioneer over the OpenAI Responses API but strips Pioneer's custom
response fields (``pioneer_savings`` / ``pioneer_routed_model``) before writing
its on-disk session rollout, so — unlike the Claude Code hook — this hook cannot
compute savings from the transcript. Instead Pioneer accumulates each turn's
savings server-side, keyed by the ``prompt_cache_key`` Codex sends on every
request (its session id). This hook derives that same session id from the
rollout and queries the cumulative figure, then prints it once per change.

For a *direct* (non-auto) model it nudges toward pioneer/auto once per session.
No-op on any error, when the API key is unknown, or when there is nothing to
show — a Stop hook must never disrupt the session.
"""

from __future__ import annotations

import json
import os
import re
import sys
import urllib.error
import urllib.request
from pathlib import Path
```

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

### Start NER Data Generation Job

Source: https://docs.pioneer.ai/api-reference/synthetic-data

Initiates an asynchronous job to generate synthetic NER examples. Specify the dataset name, labels, number of examples, and a domain description for relevance. The response provides a job ID for status polling.

```bash
curl -X POST https://api.pioneer.ai/generate \
  -H "X-API-Key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "task_type": "ner",
    "dataset_name": "my-ner-dataset",
    "labels": ["person", "company", "product"],
    "num_examples": 100,
    "domain_description": "Tech industry news articles"
  }'
```

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

### Start OpenCode CLI with Web Search

Source: https://docs.pioneer.ai/opencode

Initializes the OpenCode CLI with the Exa web search feature enabled.

```bash
OPENCODE_ENABLE_EXA=1 opencode
```

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

### Start an Evaluation Run

Source: https://docs.pioneer.ai/api-reference/evaluations

Use this endpoint to start an evaluation run. It requires the model to evaluate and the dataset name. The response includes an evaluation ID for polling results.

```bash
curl -X POST https://api.pioneer.ai/felix/evaluations \
  -H "X-API-Key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{ \
    "base_model": "YOUR_TRAINING_JOB_ID", \
    "dataset_name": "YOUR_DATASET_NAME" \
  }'
```

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

### Automate Environment Setup

Source: https://docs.pioneer.ai/api-reference/coding-agent-integration

Shell commands to create the configuration file and source it in the user's shell profile.

```bash
mkdir -p ~/.pioneer
cat > ~/.pioneer/env <<'EOF'
unset ANTHROPIC_AUTH_TOKEN CLAUDE_CODE_OAUTH_TOKEN
export ANTHROPIC_API_KEY="<YOUR_PIONEER_API_KEY>"
export ANTHROPIC_BASE_URL="https://api.pioneer.ai"
export CLAUDE_CODE_ENABLE_GATEWAY_MODEL_DISCOVERY=1
export ANTHROPIC_CUSTOM_MODEL_OPTION="pioneer/auto"
EOF
chmod 600 ~/.pioneer/env
RC=~/.zshrc
[ -n "$BASH_VERSION" ] && RC=~/.bashrc
grep -qsF '. ~/.pioneer/env' "$RC" || printf '\n# Pioneer inference\n. ~/.pioneer/env\n' >> "$RC"
. ~/.pioneer/env
```

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

### Start a Training Job

Source: https://docs.pioneer.ai/guides/agent-skills

Initiates a new training job with specified model parameters and dataset.

```bash
curl -X POST https://api.pioneer.ai/felix/training-jobs \
 -H "X-API-Key: YOUR_API_KEY" \
 -H "Content-Type: application/json" \
 -d '{
 "model_name": "my-ner-model",
 "base_model": "fastino/gliner2-base-v1",
 "datasets": [{"name": "my-dataset"}],
 "training_type": "lora",
 "nr_epochs": 5,
 "learning_rate": 5e-5
 }'
```

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

### Verify Setup with Read-Only Commands

Source: https://docs.pioneer.ai/CLI-Installation

Run diagnostic commands to ensure the CLI is correctly configured and connected.

```bash
pioneer --version
pioneer auth status
pioneer model base-models
pioneer dataset list
```

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

### Verify OpenClaw installation

Source: https://docs.pioneer.ai/openclaw

Check the installed version of OpenClaw to ensure it is available.

```shellscript
openclaw --version
```

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

### Run an Evaluation

Source: https://docs.pioneer.ai/guides/agent-skills

Starts an evaluation process for a specific model against a dataset.

```bash
curl -X POST https://api.pioneer.ai/felix/evaluations \
 -H "X-API-Key: YOUR_API_KEY" \
 -H "Content-Type: application/json" \
 -d '{
 "base_model": "job_abc123",
 "dataset_name": "my-eval-dataset"
 }'
```

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

### Install Routed Savings Hook

Source: https://docs.pioneer.ai/api-reference/coding-agent-integration

Shell script to install a hook that calculates and displays cumulative savings from the Pioneer router.

```bash
mkdir -p ~/.pioneer/hooks
cat > ~/.pioneer/hooks/show-pioneer-routed-model.sh <<'PIONEER_ROUTED_MODEL_HOOK'
#!/usr/bin/env bash
# Claude Code hook: surface how much pioneer/auto's routing has saved this session.
#
# Instead of the raw backend model, this shows cumulative money saved: for each
# pioneer/auto turn the backend stamps a per-1M-token savings rate diff vs a
# frontier reference (pioneer_savings) on the response; this hook multiplies it
# by the per-turn token usage Claude Code records and sums across the session.
# On a turn where the routed model changed (cold prompt cache), cache-write
# savings are dropped so the figure stays honest. For a *direct* (non-auto)
# model it keeps nudging toward pioneer/auto via X-Pioneer-Router-Tip.
# No-op when not on a Pioneer gateway, when no signal is present, or when
# cumulative savings are not positive.

set -euo pipefail

INPUT_FILE=$(mktemp)
trap 'rm -f "$INPUT_FILE"' EXIT
cat > "$INPUT_FILE"

python3 - "$INPUT_FILE" <<'PY' 2>/dev/null || true
from __future__ import annotations

import json
import os
import re
import sys
import time
from pathlib import Path


ANSI_RE = re.compile(r"\x1b\[[0-9;]*m")
SET_MODEL_RE = re.compile(r"Set model to\s+(.+?)\s+and saved as your default")
AUTO_ROUTER_ALIASES = {
    "pioneer/auto",
    "anthropic/pioneer-auto",
    "anthropic/pioneer/auto",
}
GATEWAY_ALIAS_PREFIX = "anthropic/pioneer/"
FRONTIER_REFERENCE_MODEL = "claude-opus-4-7"
DEFAULT_ROUTER_TIP = (
    "Tip: use model=pioneer/auto to let Pioneer route each request automatically; "
    "named models pin that concrete model."
)
TOKENS_PER_MILLION = 1_000_000.0
USAGE_TOKEN_KEYS = (
    "input_tokens",
    "output_tokens",
    "cache_read_input_tokens",
    "cache_creation_input_tokens",
    "cache_read_tokens",
    "cache_write_tokens",
)


def collect_pioneer_signals(
    value: object, routed_models: list[str], router_tips: list[str]
) -> None:
    if isinstance(value, dict):
        for key, nested in value.items():
            key_normalized = str(key).lower().replace("_", "-")
            if key_normalized == "pioneer-routed-model":
                if isinstance(nested, str) and nested:
                    routed_models.append(nested)
                continue
            if key_normalized == "x-pioneer-router-tip":
                if isinstance(nested, str) and nested:
                    router_tips.append(nested)
                continue
            collect_pioneer_signals(nested, routed_models, router_tips)
    elif isinstance(value, list):
        for item in value:
            collect_pioneer_signals(item, routed_models, router_tips)
```

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

### POST /felix/training-jobs

Source: https://docs.pioneer.ai/api-reference/overview

Starts a new model training job.

```APIDOC
## POST /felix/training-jobs

### Description
Starts a training job using specified model name, datasets, and base model.

### Method
POST

### Endpoint
/felix/training-jobs
```

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

### Start Claude Code

Source: https://docs.pioneer.ai/api-reference/coding agent integration

Initialize the Claude Code agent after setting the environment variables.

```bash
claude
```

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

### Verify Hermes Setup

Source: https://docs.pioneer.ai/hermes

Perform a non-interactive check to confirm Hermes is responding correctly.

```bash
hermes chat -q "Reply with exactly PIONEER_HERMES_OK"
```

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

### Codex Setup Prompt

Source: https://docs.pioneer.ai/codex

A prompt to be entered into Codex to automate the configuration of the local model catalog and environment settings.

```text
Set up this machine's Codex CLI to use Pioneer, including a fresh local model catalog.
  Do exactly this, in order:
  1. Read my Pioneer API key from the PIONEER_API_KEY environment variable. If it is not set, stop and ask me for it — never
  guess or hardcode a key, and never print the key value.
  2. Determine this user's absolute home directory (e.g. run `echo "$HOME"`). Call it HOME_ABS. Everywhere below, use the
  literal absolute path — do NOT write "~" or "$HOME" into any file, because Codex does not expand them in config.toml.
  3. Fetch the Pioneer model list:
     curl -fsS https://api.pioneer.ai/v1/models -H "Authorization: Bearer $PIONEER_API_KEY"
     If the request fails (non-200), show me the status and body and STOP without changing any files.
  4. From the JSON response, take ONLY the top-level "models" array (not "data"). Wrap it as {"models": [ ... ]} and write it
  pretty-printed to:
     HOME_ABS/.codex/model-catalogs/pioneer.json
     Create HOME_ABS/.codex/model-catalogs/ if needed. Overwrite the file if it exists.
  5. Back up HOME_ABS/.codex/config.toml to config.toml.bak (skip if config.toml doesn't exist yet). Then ensure config.toml
  has exactly these Pioneer settings, written as TOP-LEVEL keys plus the [model_providers.pioneer] table (substitute HOME_ABS
  into the catalog path):
     model = "pioneer/auto"
     model_provider = "pioneer"
     model_reasoning_effort = "medium"
     model_catalog_json = "HOME_ABS/.codex/model-catalogs/pioneer.json"
     [model_providers.pioneer]
     name = "Pioneer"
     base_url = "https://api.pioneer.ai/v1"
     wire_api = "responses"
     supports_websockets = false
     Preserve any unrelated existing sections (e.g. [projects.*], [tui.*], other providers). Only set/replace the keys above
  and the [model_providers.pioneer] table. Make sure model_catalog_json is a top-level key, NOT nested inside
  [model_providers.pioneer].
  6. Report how many models were written to the catalog, confirm the config keys are set, and remind me to restart Codex so
  the /model picker reloads.
```

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

### List all datasets

Source: https://docs.pioneer.ai/api-reference/datasets

Retrieves a list of all datasets associated with the account using a GET request.

```bash
curl https://api.pioneer.ai/felix/datasets \
  -H "X-API-Key: YOUR_API_KEY"
```

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

### GET /training-jobs/{job_id}/download

Source: https://docs.pioneer.ai/concepts/training

Requests a presigned URL to download model weights.

```APIDOC
## GET /training-jobs/{job_id}/download

### Description
Requests a presigned URL to download model weights. Requires a Pro plan or above.

### Method
GET

### Endpoint
https://api.pioneer.ai/felix/training-jobs/{job_id}/download

### Parameters
#### Path Parameters
- **job_id** (string) - Required - The unique identifier of the training job.

### Response
#### Success Response (200)
- **success** (boolean) - Indicates if the request was successful.
- **download_url** (string) - The presigned URL for downloading the weights.
- **expires_in_seconds** (integer) - Expiration time for the URL in seconds.
- **file_name** (string) - The name of the file being downloaded.
```

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

### Launch Claude Code

Source: https://docs.pioneer.ai/api-reference/coding-agent-integration

Starts the Claude Code agent using the Pioneer auto-router model.

```bash
claude --model pioneer/auto
```

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

### Get a presigned upload URL

Source: https://docs.pioneer.ai/concepts/datasets

Request a presigned URL to initiate the upload process. Only dataset_name is required.

```bash
curl -X POST https://api.pioneer.ai/felix/datasets/upload/url \
    -H "X-API-Key: YOUR_API_KEY" \
    -H "Content-Type: application/json" \
    -d '{
      "dataset_name": "my-ner-dataset",
      "dataset_type": "ner",
      "type": "training",
      "filename": "data.jsonl"
    }'
```

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

### Pioneer Stop Hook Output Example

Source: https://docs.pioneer.ai/api-reference/coding-agent-integration

Example of the informational message displayed by the Stop hook after a turn, indicating cumulative session savings.

```text
Stop says: Pioneer routing saved ~$1.43 this session (vs claude-opus-4-7)
```

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

### Test connectivity with a placeholder key

Source: https://docs.pioneer.ai/api-reference/authentication

Verify network reachability by sending a request with a test key to confirm the integration is wired correctly.

```bash
curl -X POST https://api.pioneer.ai/v1/messages \
  -H "X-API-Key: pio_sk_test" \
  -H "Content-Type: application/json" \
  -d '{"model":"claude-haiku-5","max_tokens":10,"messages":[{"role":"user","content":"hi"}]}'

# Expected: {"detail":"Invalid API key format. API keys must start with 'pio_sk_'. Please check your X-API-Key header."}
# A 401 with this body = integration is wired correctly. Swap in a real key to get completions.
```

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

### Download Model Weights

Source: https://docs.pioneer.ai/concepts/training

Request a presigned URL to download model weights. Requires a Pro plan or higher.

```bash
curl https://api.pioneer.ai/felix/training-jobs/3fa85f64-5717-4562-b3fc-2c963f66afa6/download \
  -H "X-API-Key: YOUR_API_KEY"
```

```json
{
  "success": true,
  "download_url": "https://...",
  "expires_in_seconds": 3600,
  "file_name": "my-ner-model-weights.zip"
}
```

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

### Configure and Run Codex

Source: https://docs.pioneer.ai/api-reference/coding agent integration

Point the Codex CLI to Pioneer by providing the API key and provider configuration flags.

```bash
PIONEER_API_KEY="<Your Pioneer API key>" codex \
    -c 'model_provider="pioneer"' \
    -c 'model_providers.pioneer.name="Pioneer"' \
    -c 'model_providers.pioneer.base_url="https://api.pioneer.ai/v1"' \
    -c 'model_providers.pioneer.wire_api="responses"' \
    -c 'model_providers.pioneer.env_key="PIONEER_API_KEY"'
```

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

### Create a project

Source: https://docs.pioneer.ai/api-reference/projects

Creates a new project.

```APIDOC
## Create a project

`POST /projects`

### Description
Creates a new project.

### Method
POST

### Endpoint
/projects

### Parameters
#### Request Body
- **name** (string) - Required - A unique, human-readable name for the project.

### Request Example
```bash
curl -X POST https://api.pioneer.ai/projects \
  -H "X-API-Key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "production-ner"
  }'
```

### Response
#### Success Response (200)
- **id** (string) - The new project's unique ID. Use this in all subsequent project and deployment requests.
- **name** (string) - The project name.
```

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

### GET /v1/models/:model_id

Source: https://docs.pioneer.ai/concepts/inference

Retrieves metadata for a specific model.

```APIDOC
## GET /v1/models/:model_id

### Description
Retrieve a single model's metadata.

### Method
GET

### Endpoint
https://api.pioneer.ai/v1/models/:model_id

### Parameters
#### Path Parameters
- **model_id** (string) - Required - The unique identifier of the model.
```

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

### Mintlify development commands

Source: https://docs.pioneer.ai/AGENTS

Commands for previewing the documentation site and validating links.

```bash
mint dev
```

```bash
mint broken-links
```

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

### GET /v1/models

Source: https://docs.pioneer.ai/concepts/inference

Lists all available models in the public catalog.

```APIDOC
## GET /v1/models

### Description
Lists available models (callable without auth for the public catalog).

### Method
GET

### Endpoint
https://api.pioneer.ai/v1/models
```

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

### List Projects

Source: https://docs.pioneer.ai/api-reference/projects

Fetches all projects associated with your account. Requires your API key for authentication.

```bash
curl https://api.pioneer.ai/projects \
  -H "X-API-Key: YOUR_API_KEY"
```

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

### GET /felix/training-jobs/:id

Source: https://docs.pioneer.ai/api-reference/overview

Polls the status of a training job.

```APIDOC
## GET /felix/training-jobs/:id

### Description
Polls the training job status until 'complete'. Returns performance metrics upon completion.

### Method
GET

### Endpoint
/felix/training-jobs/:id

### Parameters
#### Path Parameters
- **id** (string) - Required - The training job ID.
```

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

### Initialize Local Gateway and Dashboard

Source: https://docs.pioneer.ai/openclaw

Configures the gateway mode, generates an authentication token if missing, and restarts the service to apply changes.

```shell
openclaw config set gateway.mode local
GATEWAY_TOKEN="$(openclaw config get gateway.auth.token 2>/dev/null || true)"
if [ -z "$GATEWAY_TOKEN" ] || [ "$GATEWAY_TOKEN" = "null" ]; then
  GATEWAY_TOKEN="$(openssl rand -hex 32)"
  openclaw config set gateway.auth.token "$GATEWAY_TOKEN"
fi
openclaw gateway install --force
openclaw gateway restart
openclaw dashboard
unset GATEWAY_TOKEN
```

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

### GET /base-models

Source: https://docs.pioneer.ai/guides/fine-tune-llm

Retrieve a list of base models available for training.

```APIDOC
## GET /base-models

### Description
Retrieves the catalog of base models that support training.

### Method
GET

### Endpoint
https://api.pioneer.ai/base-models

### Query Parameters
- **task_type** (string) - Optional - The task type (e.g., 'decoder').
- **supports_training** (boolean) - Optional - Filter for models that support training.
```

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

### Authenticate with Pioneer

Source: https://docs.pioneer.ai/CLI-Installation

Log in interactively and check the current authentication status.

```bash
pioneer auth login
```

```bash
pioneer auth status
```

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

### Create Project

Source: https://docs.pioneer.ai/api-reference/projects

Creates a new project with a specified name. The response includes the new project's ID, which is essential for subsequent requests.

```bash
curl -X POST https://api.pioneer.ai/projects \
  -H "X-API-Key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "production-ner"
  }'
```

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

### GET /felix/evaluations/:id

Source: https://docs.pioneer.ai/guides/fine-tune-ner

Retrieve the results of a specific evaluation job.

```APIDOC
## GET /felix/evaluations/:id

### Description
Retrieve evaluation results. Results include f1, precision, recall, and a per_entity breakdown.

### Method
GET

### Endpoint
https://api.pioneer.ai/felix/evaluations/:id
```

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

### Submit a training job via cURL

Source: https://docs.pioneer.ai/api-reference/training-jobs

Examples for initiating fine-tuning jobs for NER and LLM models. Ensure the X-API-Key header is set and the base_model ID is valid.

```bash
curl -X POST https://api.pioneer.ai/felix/training-jobs \
  -H "X-API-Key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model_name": "my-ner-model",
    "base_model": "fastino/gliner2-base-v1",
    "datasets": [{"name": "my-dataset"}],
    "training_type": "lora",
    "nr_epochs": 5,
    "learning_rate": 5e-5
  }'
```

```bash
curl -X POST https://api.pioneer.ai/felix/training-jobs \
  -H "X-API-Key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model_name": "my-sft-model",
    "base_model": "nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16",
    "datasets": [{"name": "my-dataset"}],
    "training_type": "lora"
  }'
```

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

### GET /inferences

Source: https://docs.pioneer.ai/concepts/inference

Retrieve a list of recent inference calls with optional filtering.

```APIDOC
## GET /inferences

### Description
List recent inference records. Supports filtering by various parameters.

### Method
GET

### Endpoint
https://api.pioneer.ai/inferences

### Parameters
#### Query Parameters
- **limit** (integer) - Optional
- **offset** (integer) - Optional
- **model_id** (string) - Optional
- **task** (string) - Optional
- **project_id** (string) - Optional
- **training_job_id** (string) - Optional
- **latency_min** (integer) - Optional - Minimum latency in ms
- **latency_max** (integer) - Optional - Maximum latency in ms
- **since** (string) - Optional - ISO 8601 start bound
- **until** (string) - Optional - ISO 8601 end bound
```

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

### GET /felix/datasets/{dataset_name}

Source: https://docs.pioneer.ai/guides/fine-tune-classification

Retrieve information about a specific dataset to check its status.

```APIDOC
## GET /felix/datasets/{dataset_name}

### Description
Retrieve details about a dataset, including its current status.

### Method
GET

### Endpoint
https://api.pioneer.ai/felix/datasets/{dataset_name}

### Parameters
#### Path Parameters
- **dataset_name** (string) - Required - The name of the dataset to retrieve.
```

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

### List training-compatible base models

Source: https://docs.pioneer.ai/guides/fine-tune-llm

Retrieve a list of decoder models that support training.

```bash
curl "https://api.pioneer.ai/base-models?task_type=decoder&supports_training=true" \
  -H "X-API-Key: YOUR_API_KEY"
```

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

### GET /felix/datasets/:name

Source: https://docs.pioneer.ai/api-reference/datasets

Retrieves the version history and metadata for a specific dataset.

```APIDOC
## GET /felix/datasets/:name

### Description
Returns version history and metadata for a specific dataset.

### Method
GET

### Endpoint
/felix/datasets/:name

### Parameters
#### Path Parameters
- **name** (string) - Required - The dataset name exactly as it appears in your account.

### Response
#### Success Response (200)
- **success** (boolean) - true on success.
- **count** (number) - Number of versions returned.
- **versions** (object[]) - List of dataset versions, most recent first.

#### Response Example
{
  "success": true,
  "count": 1,
  "versions": [
    {
      "id": "v123",
      "dataset_name": "my-ner-dataset",
      "version_number": "1",
      "size": 100,
      "status": "ready",
      "created_at": "2023-10-27T10:00:00Z"
    }
  ]
}
```

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

### Run a Model Evaluation

Source: https://docs.pioneer.ai/concepts/evaluations

Use this command to initiate an evaluation for a fine-tuned model. Pass your training job ID as `base_model` and the name of your evaluation dataset. Ensure you replace 'YOUR_API_KEY' with your actual API key.

```bash
curl -X POST https://api.pioneer.ai/felix/evaluations \
  -H "X-API-Key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "base_model": "job_abc123",
    "dataset_name": "my-eval-dataset"
  }'
```

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

### GET /felix/datasets

Source: https://docs.pioneer.ai/api-reference/datasets

Retrieves a list of all datasets associated with the user's account.

```APIDOC
## GET /felix/datasets

### Description
Returns all datasets associated with your account.

### Method
GET

### Endpoint
/felix/datasets

### Response
#### Success Response (200)
- **success** (boolean) - true on success.
- **count** (number) - Total number of datasets returned.
- **datasets** (object[]) - Array of dataset objects.

#### Response Example
{
  "success": true,
  "count": 1,
  "datasets": [
    {
      "dataset_name": "my-ner-dataset",
      "status": "ready",
      "created_at": "2023-10-27T10:00:00Z",
      "version_number": "1"
    }
  ]
}
```

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

### GET /felix/datasets/{name}/{version}

Source: https://docs.pioneer.ai/api-reference/overview

Polls the status of a dataset processing job.

```APIDOC
## GET /felix/datasets/{name}/{version}

### Description
Polls the status of a dataset until it is 'ready'.

### Method
GET

### Endpoint
/felix/datasets/{name}/{version}

### Parameters
#### Path Parameters
- **name** (string) - Required - The name of the dataset.
- **version** (string) - Required - The version of the dataset.
```

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

### GET /generate/jobs/{job_id}

Source: https://docs.pioneer.ai/api-reference/overview

Polls the status of a synthetic data generation job.

```APIDOC
## GET /generate/jobs/{job_id}

### Description
Polls the status of a synthetic data generation job until it returns 'ready'.

### Method
GET

### Endpoint
/generate/jobs/{job_id}

### Parameters
#### Path Parameters
- **job_id** (string) - Required - The unique identifier for the generation job.
```

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

### Example Savings Output

Source: https://docs.pioneer.ai/api-reference/coding-agent-integration

The format of the savings message displayed by the hook at the end of a turn.

```text
Pioneer routing saved ~$1.43 this session (vs claude-opus-4-7)
```

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

### Synthetic Data API

Source: https://docs.pioneer.ai/llms.txt

Endpoints for starting data generation jobs and auto-labeling text.

```APIDOC
## POST /generate

### Description
Starts a Pioneer data generation job for NER, classification, or decoder tasks.

### Method
POST

### Endpoint
/generate
```

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

### Configure Hook and Settings

Source: https://docs.pioneer.ai/api-reference/coding-agent-integration

Appends a stop command hook and writes the configuration settings to the specified path.

```python
hooks.setdefault("Stop", []).append(
    {"hooks": [{"type": "command", "command": command, "timeout": 5}]}
)
settings_path.parent.mkdir(parents=True, exist_ok=True)
settings_path.write_text(json.dumps(settings, indent=2) + "\n")
```

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

### Text completions

Source: https://docs.pioneer.ai/api-reference/inference/openai-compatible

The `/v1/completions` endpoint supports the legacy completions format with a `prompt` field. Similar to chat completions, use your training job ID as the `model` and include Pioneer-specific fields via `extra_body`.

```APIDOC
## Text completions

`POST /v1/completions` supports the legacy completions format with a `prompt` field.

```python
response = client.completions.create(
    model="YOUR_TRAINING_JOB_ID",
    prompt="Extract the company names from: Apple and Google announced a partnership.",
    extra_body={
        "schema": {
            "entities": ["organization"]
        }
    }
)

print(response.choices[0].text)
```

```bash
curl -X POST https://api.pioneer.ai/v1/completions \
  -H "X-API-Key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d 
    "{
      \"model\": \"YOUR_TRAINING_JOB_ID\",
      \"prompt\": \"Extract the company names from: Apple and Google announced a partnership.\",
      \"schema\": {
        \"entities\": [\"organization\"]
      }
    }"
```
```

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

### GET /inferences/{INFERENCE_ID}/feedback

Source: https://docs.pioneer.ai/concepts/inference

Read back feedback previously submitted for an inference.

```APIDOC
## GET /inferences/{INFERENCE_ID}/feedback

### Description
Retrieve feedback submitted for a specific inference.

### Method
GET

### Endpoint
https://api.pioneer.ai/inferences/{INFERENCE_ID}/feedback

### Parameters
#### Path Parameters
- **INFERENCE_ID** (string) - Required
```

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

### Submit a training job

Source: https://docs.pioneer.ai/guides/fine-tune-ner

Initiate a LoRA fine-tuning job using a specified base model and dataset.

```bash
curl -X POST https://api.pioneer.ai/felix/training-jobs \
  -H "X-API-Key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model_name": "my-ner-model",
    "base_model": "fastino/gliner2-base-v1",
    "datasets": [{"name": "my-ner-dataset"}],
    "training_type": "lora",
    "nr_epochs": 5,
    "learning_rate": 5e-5
  }'
```

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

### List and Deploy Checkpoints

Source: https://docs.pioneer.ai/concepts/training

Retrieve a list of available checkpoints for a job or deploy a specific checkpoint to an inference endpoint.

```bash
curl https://api.pioneer.ai/felix/training-jobs/3fa85f64-5717-4562-b3fc-2c963f66afa6/checkpoints \
  -H "X-API-Key: YOUR_API_KEY"
```

```bash
curl -X POST https://api.pioneer.ai/felix/training-jobs/3fa85f64-5717-4562-b3fc-2c963f66afa6/checkpoints/CHECKPOINT_ID/deploy \
  -H "X-API-Key: YOUR_API_KEY"
```

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

### GET /training-jobs/{job_id}/checkpoints

Source: https://docs.pioneer.ai/concepts/training

Lists all checkpoints associated with a specific training job.

```APIDOC
## GET /training-jobs/{job_id}/checkpoints

### Description
Retrieves a list of checkpoints for the specified training job. Each checkpoint includes flags for 'is_best', 'is_final', and 'is_deployable'.

### Method
GET

### Endpoint
https://api.pioneer.ai/felix/training-jobs/{job_id}/checkpoints

### Parameters
#### Path Parameters
- **job_id** (string) - Required - The unique identifier of the training job.
```

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

### GET /felix/datasets/{name}/{version}

Source: https://docs.pioneer.ai/concepts/datasets

Retrieves the status and details of a specific dataset version.

```APIDOC
## GET /felix/datasets/{name}/{version}

### Description
Fetches the current status of a dataset. Poll this endpoint until the status is 'ready'.

### Method
GET

### Endpoint
/felix/datasets/{name}/{version}

### Parameters
#### Path Parameters
- **name** (string) - Required - The name of the dataset.
- **version** (string) - Required - The version number or 'latest'.
```

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

### List available models with OpenAI SDK and Pioneer

Source: https://docs.pioneer.ai/api-reference/inference/openai-compatible

Retrieve a list of available models using the OpenAI SDK by calling the `models.list()` method.

```python
models = client.models.list()
for model in models.data:
    print(model.id)
```

```bash
curl https://api.pioneer.ai/v1/models \
  -H "X-API-Key: YOUR_API_KEY"
```

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

### List projects

Source: https://docs.pioneer.ai/api-reference/projects

Returns all projects in your account.

```APIDOC
## List projects

`GET /projects`

### Description
Returns all projects in your account.

### Method
GET

### Endpoint
/projects

### Response
#### Success Response (200)
- **projects** (object[]) - Array of project objects.
  - **id** (string) - Unique project ID.
  - **name** (string) - Human-readable project name.
  - **created_at** (string) - ISO 8601 creation timestamp.

### Request Example
```bash
curl https://api.pioneer.ai/projects \
  -H "X-API-Key: YOUR_API_KEY"
```
```

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

### Get evaluation results

Source: https://docs.pioneer.ai/api-reference/evaluations

Returns the status and, once complete, the full results of an evaluation run.

```APIDOC
## Get evaluation results

### Description
Returns the status and, once complete, the full results of an evaluation run.

### Method
GET

### Endpoint
/felix/evaluations/:id

### Parameters
#### Path Parameters
- **id** (string) - Required - The evaluation UUID.

### Request Example
```bash
curl https://api.pioneer.ai/felix/evaluations/YOUR_EVALUATION_ID \
  -H "X-API-Key: YOUR_API_KEY"
```

### Response
#### Success Response (200)
- **id** (string) - Evaluation UUID.
- **status** (string) - Current status of the evaluation. Values: `queued`, `running`, `complete`, `failed`.
- **f1_score** (number) - Overall F1 score. Present once the evaluation is complete.
- **precision_score** (number) - Overall precision score.
- **recall_score** (number) - Overall recall score.
- **sample_count** (number) - Number of examples evaluated.
- **completed_at** (string) - ISO 8601 timestamp of when the evaluation finished.

#### Response Example
{
  "id": "YOUR_EVALUATION_ID",
  "status": "complete",
  "f1_score": 0.85,
  "precision_score": 0.88,
  "recall_score": 0.82,
  "sample_count": 1000,
  "completed_at": "2023-10-27T10:00:00Z"
}
```

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

### Show Hermes Configuration

Source: https://docs.pioneer.ai/hermes

Display the current active provider and configuration settings.

```bash
hermes config show
```

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

### Troubleshoot CLI Path

Source: https://docs.pioneer.ai/CLI-Installation

Check the global npm binary directory to ensure it is in the system PATH.

```bash
npm bin -g
```

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

### GET /felix/trained-models

Source: https://docs.pioneer.ai/api-reference/training-jobs

Returns a flat list of all successfully trained models across all training jobs.

```APIDOC
## GET /felix/trained-models

### Description
Returns a flat list of all successfully trained models across all of your training jobs.

### Method
GET

### Endpoint
/felix/trained-models
```

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

### Perform inference with compatible SDKs

Source: https://docs.pioneer.ai/quickstart

Use Pioneer's endpoints with existing OpenAI or Anthropic SDKs by pointing to the Pioneer base URL.

```bash
curl -X POST https://api.pioneer.ai/v1/chat/completions \
  -H "X-API-Key: $PIONEER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "fastino/gliner2-base-v1",
    "messages": [
      {"role": "user", "content": "Extract entities from: Apple launched the iPhone in San Francisco."}
    ],
    "schema": {"entities": ["organization", "product", "location"]}
  }'
```

```bash
curl -X POST https://api.pioneer.ai/v1/messages \
  -H "X-API-Key: $PIONEER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "fastino/gliner2-base-v1",
    "max_tokens": 1024,
    "messages": [
      {"role": "user", "content": "Extract entities from: Apple launched the iPhone in San Francisco."}
    ],
    "schema": {"entities": ["organization", "product", "location"]}
  }'
```

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

### GET /felix/training-jobs/:id/download

Source: https://docs.pioneer.ai/api-reference/training-jobs

Provides a download URL for trained model weights once a job is complete.

```APIDOC
## GET /felix/training-jobs/:id/download

### Description
Returns a download URL for the trained model weights. Only available once the job status is complete.

### Method
GET

### Endpoint
/felix/training-jobs/:id/download

### Parameters
#### Path Parameters
- **id** (string) - Required - The training job UUID.
```

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

### GET /base-models

Source: https://docs.pioneer.ai/concepts/models

Retrieves the live catalog of base models, including availability for training or inference.

```APIDOC
## GET /base-models

### Description
Retrieves the live catalog of base models. Use this to check for availability, context limits, and pricing before submitting training jobs or inference requests.

### Method
GET

### Endpoint
/base-models

### Parameters
#### Query Parameters
- **supports_training** (boolean) - Optional - Filter models that support LoRA fine-tuning.
- **supports_inference** (boolean) - Optional - Filter models that support serverless inference.
```
Tokens:
Raw