morphik.toml

[api]
host = "0.0.0.0"
port = 8000
reload = true

[service]
environment = "development"  # "development", "staging", or "production"
version = "unknown"
enable_profiling = false  # Enable request profiling (creates .prof files in logs/)

[auth]
jwt_algorithm = "HS256"
bypass_auth_mode = true  # Enabled by default for easier local development
dev_user_id = "dev_user"  # Default dev user ID

#### Registered models
[registered_models]
# OpenAI models
openai_gpt4-1 = { model_name = "gpt-4.1" }
openai_gpt4-1-mini = { model_name = "gpt-4.1-mini" }

# Azure OpenAI models
azure_gpt4 = { model_name = "gpt-4", api_base = "YOUR_AZURE_URL_HERE", api_version = "2023-05-15", deployment_id = "gpt-4-deployment" }
azure_gpt35 = { model_name = "gpt-3.5-turbo", api_base = "YOUR_AZURE_URL_HERE", api_version = "2023-05-15", deployment_id = "gpt-35-turbo-deployment" }

# Anthropic models
claude_opus = { model_name = "claude-3-opus-20240229" }
claude_sonnet = { model_name = "claude-3-7-sonnet-latest" }

# Google Gemini models
gemini_flash = { model_name = "gemini/gemini-2.5-flash-preview-05-20" } # gemini-2.5-flash-preview-05-20
# gemini_flash = { model_name = "gemini/gemini-2.5-pro-preview-06-05" } # { model_name = "claude-4-sonnet-20250514"} # { model_name = "gpt-4.1" } # { model_name = "gemini/gemini-2.5-pro-preview-06-05" } # {model_name = "o3-2025-04-16"} #  { model_name = "claude-3-7-sonnet-latest"} #  { model_name = "gemini/gemini-2.5-flash-preview-05-20" } # gemini-2.5-flash-preview-05-20
# gemini_flash = { model_name = "groq/meta-llama/llama-4-maverick-17b-128e-instruct"}

# Ollama models (modify api_base based on your deployment)
# - Local Ollama: "http://localhost:11434" (default)
# - Morphik in Docker, Ollama local: "http://host.docker.internal:11434"
# - Both in Docker: "http://ollama:11434"

# ollama_qwen_vision = { model_name = "gpt-4.1" } # {model_name = "groq/meta-llama/llama-4-scout-17b-16e-instruct"} #{ model_name = "ollama_chat/qwen2.5vl:latest", api_base = "http://localhost:11434", vision = true }
ollama_qwen_vision = { model_name = "ollama_chat/qwen2.5vl:latest", api_base = "http://localhost:11434", vision = true }

ollama_llama_vision = { model_name = "ollama_chat/llama3.2-vision", api_base = "http://localhost:11434", vision = true }
ollama_qwen_1_5b = { model_name = "ollama_chat/qwen2:1.5b", api_base = "http://ollama:11434" }

ollama_embedding = { model_name = "ollama/nomic-embed-text", api_base = "http://localhost:11434" }

# Lemonade models (for AMD GPU and NPU support)
# - Local: "http://localhost:8020/api/v1"
# - Morphik in Docker: "http://host.docker.internal:8020/api/v1"
# lemonade_qwen = { model_name = "openai/Qwen2.5-VL-7B-Instruct-GGUF", api_base = "http://localhost:8020/api/v1", vision = true }
lemonade_embedding = { model_name = "openai/nomic-embed-text-v1-GGUF", api_base = "http://localhost:8020/api/v1" }

openai_embedding = { model_name = "text-embedding-3-small" }
openai_embedding_large = { model_name = "text-embedding-3-large" }
azure_embedding = { model_name = "text-embedding-ada-002", api_base = "YOUR_AZURE_URL_HERE", api_version = "2023-05-15", deployment_id = "embedding-ada-002" }


#### Component configurations ####

[completion]
model = "ollama_qwen_vision" #"openai_gpt4-1-mini"  # Reference to a key in registered_models
default_max_tokens = "1000"
default_temperature = 0.3

[database]
provider = "postgres"
# Connection pool settings
pool_size = 10           # Maximum number of connections in the pool
max_overflow = 15        # Maximum number of connections that can be created beyond pool_size
pool_recycle = 3600      # Time in seconds after which a connection is recycled (1 hour)
pool_timeout = 10        # Seconds to wait for a connection from the pool
pool_pre_ping = true     # Check connection viability before using it from the pool
max_retries = 3          # Number of retries for database operations
retry_delay = 1.0        # Initial delay between retries in seconds

[embedding]
model = "ollama_embedding"  # Reference to registered model
dimensions = 768
similarity_metric = "cosine"

[parser]
chunk_size = 6000
chunk_overlap = 300
use_contextual_chunking = false
contextual_chunking_model = "ollama_qwen_vision"  # Reference to a key in registered_models

[parser.xml]
max_tokens = 350
preferred_unit_tags = ["SECTION", "Section", "Article", "clause"]
ignore_tags = ["TOC", "INDEX"]

[document_analysis]
model = "ollama_qwen_vision"  # Reference to a key in registered_models

[parser.vision]
model = "ollama_qwen_vision"  # Reference to a key in registered_models
frame_sample_rate = -1  # Set to -1 to disable frame captioning

[reranker]
use_reranker = false
provider = "flag"
model_name = "BAAI/bge-reranker-large"
query_max_length = 256
passage_max_length = 512
use_fp16 = true
device = "mps" # use "cpu" if on docker and using a mac, "cuda" if cuda enabled device

[storage]
provider = "local"
storage_path = "./storage"
cache_enabled = false
cache_max_size_gb = 10
s3_upload_concurrency = 16
# cache_path = "./storage/cache"

# [storage]
# provider = "aws-s3"
# region = "us-east-2"
# bucket_name = "morphik-s3-storage"
# storage_path = "morphik-storage"
# cache_enabled = true
# cache_max_size_gb = 10
# cache_path = "./storage/cache"
# s3_upload_concurrency = 16

[vector_store]
provider = "pgvector"

[multivector_store]
provider = "postgres"  # "morphik" # "postgres"  # "morphik" # "postgres"  # "postgres" or "morphik" for fast implementation

[redis]
url = "redis://localhost:6379/0"  # Full Redis URL (takes precedence over host/port)
host = "localhost"
port = 6379

[worker]
arq_max_jobs = 1  # Maximum concurrent jobs for ARQ worker
colpali_store_batch_size = 16  # Batch size for ColPali vector storage

[pdf]
colpali_pdf_dpi = 150  # DPI for PDF to image conversion in ColPali processing

[morphik]
enable_colpali = true
mode = "self_hosted"  # "cloud" or "self_hosted"
secret_manager = "env"  # "env" (default) or "infisical"
api_domain = "api.morphik.ai"  # API domain for cloud URIs
# ColPali API endpoints (used when colpali_mode = "api" or parser_mode = "api")
morphik_embedding_api_domain = ["http://localhost:6000"]
colpali_mode = "local" # "off", "local", or "api"
parser_mode = "local" # "local" or "api" - "api" offloads parsing to GPU servers

[pdf_viewer]
frontend_url = "http://localhost:3000/api/pdf" # "https://morphik.ai/api/pdf" # "http://localhost:3000/api/pdf" # "https://morphik.ai/api/pdf"

[telemetry]
service_name = "databridge-core"
project_name = ""
upload_interval_hours = 4.0
max_local_bytes = 1073741824