TensorZero Configs

docker-compose.yaml

services:
  clickhouse:
    image: clickhouse/clickhouse-server:25.8-alpine
    ports:
      - "8123:8123"
    environment:
      - 'CLICKHOUSE_USER=${CLICKHOUSE_USER:?Environment variable CLICKHOUSE_USER must be set.}'
      - 'CLICKHOUSE_PASSWORD=${CLICKHOUSE_PASSWORD:?Environment variable CLICKHOUSE_USER must be set.}'
  gateway:
    depends_on:
      - clickhouse 
    image: tensorzero/gateway
    volumes:
      - '/home/limcheekin/docker/volumes/tensorzero/config:/app/config:ro'
    command: '--config-file /app/config/tensorzero.toml'
    environment:
      - 'TENSORZERO_CLICKHOUSE_URL=${TENSORZERO_CLICKHOUSE_URL:?Environment variable TENSORZERO_CLICKHOUSE_URL must be set.}'
      - 'LOCALAI_API_KEY=${LOCALAI_API_KEY:-sk-1}'
      - 'OPENAI_API_KEY=${OPENAI_API_KEY:-sk-1}'
    ports:
      - '5002:3000'
    extra_hosts:
      - 'host.docker.internal:host-gateway'
  ui:
    depends_on:
      - clickhouse    
    image: tensorzero/ui
    volumes:
      - '/home/limcheekin/docker/volumes/tensorzero/config:/app/config:ro'  
    environment:
      - 'LOCALAI_API_KEY=${LOCALAI_API_KEY:-sk-1}'
      - 'TENSORZERO_CLICKHOUSE_URL=${TENSORZERO_CLICKHOUSE_URL:?Environment variable TENSORZERO_CLICKHOUSE_URL must be set.}'
      - 'TENSORZERO_GATEWAY_URL=${TENSORZERO_GATEWAY_URL:-http://gateway:3000}'
    ports:
      - '5003:4000'

tensorzero.toml

# config/tensorzero.toml

##############################
# 1. Define Model & Provider #
##############################

[models.granite_4_h_tiny]
routing = ["local_ai"]

[models.granite_4_h_tiny.providers.local_ai]
type = "openai"
api_base = "http://192.168.1.111:8880/v1"
model_name = "ibm-granite_granite-4.0-h-tiny"
api_key_location = "env::LOCALAI_API_KEY"
extra_body = [
  { pointer = "/cache_prompt", value = true}
]

# This function will be the 'model' you select in Open WebUI
###################################
# 2. Define Function & Variants  #
###################################

[functions.granite_4_h]
type = "chat"

[functions.granite_4_h.variants.local_ai]
type = "chat_completion"
model = "granite_4_h_tiny"
weight = 100

# config/tensorzero.toml
# ---------------------
# Minimal TensorZero config for calling OpenAI GPT-4.1
# Notes:
#  - Ensure OPENAI_API_KEY is set in the environment (or change api_key_location).
#  - Model section name (gpt_4_1) is used as "tensorzero::model_name::gpt_4_1".
# ---------------------

[models.gpt_4_1]
routing = ["openai"]

[models.gpt_4_1.providers.openai]
type = "openai"
# Official OpenAI model id for GPT-4.1
model_name = "gpt-4.1"
# Use environment variable OPENAI_API_KEY by default.
api_key_location = "env::OPENAI_API_KEY"
# Optional: api_base (if you use a proxy or Azure set a custom base)
# api_base = "https://api.openai.com/v1"

# A simple chat "function" that callers can invoke via the TensorZero gateway.
[functions.chat_gpt4_1]
type = "chat"
description = "Default chat function using OpenAI gpt-4.1"

[functions.chat_gpt4_1.variants.default]
type = "chat_completion"
model = "gpt_4_1"

# Default request-time overrides for this variant.
# These can still be overridden by callers (preferred).
temperature = 0.0        # keep replies focused & deterministic — tweak as needed
top_p = 0.95
max_tokens = 2048        # output token cap; tune per your app
presence_penalty = 0.0
frequency_penalty = 0.0