Configuration

micasa --demo /tmp/my-demo.db   # creates and populates
micasa /tmp/my-demo.db          # reopens with the demo data

micasa config llm.model             # current model name
micasa config llm.base_url          # LLM API endpoint
micasa config documents.max_file_size  # max doc size in bytes

micasa backup ~/backups/micasa-$(date +%F).db
micasa backup --source /path/to/micasa.db ~/backups/snapshot.db

export MICASA_DB_PATH=/path/to/my/house.db
micasa   # uses /path/to/my/house.db

export OLLAMA_HOST=http://192.168.1.50:11434
micasa   # connects to http://192.168.1.50:11434/v1

export MICASA_LLM_MODEL=llama3.3
micasa   # uses llama3.3 instead of the default qwen3

export MICASA_LLM_TIMEOUT=15s
micasa   # waits up to 15s for LLM server responses

export MICASA_MAX_DOCUMENT_SIZE="100 MiB"
micasa   # allows documents up to 100 MiB

export MICASA_CACHE_TTL=7d
micasa   # evicts cache entries older than 7 days

# micasa configuration

[llm]
# Base URL for an OpenAI-compatible API endpoint.
# Ollama (default): http://localhost:11434/v1
# llama.cpp:        http://localhost:8080/v1
# LM Studio:        http://localhost:1234/v1
base_url = "http://localhost:11434/v1"

# Model name passed in chat requests.
model = "qwen3"

# Optional: custom context appended to all system prompts.
# Use this to inject domain-specific details about your house, currency, etc.
# extra_context = "My house is a 1920s craftsman in Portland, OR. All budgets are in CAD."

# Timeout for quick LLM server operations (ping, model listing).
# Go duration syntax: "5s", "10s", "500ms", etc. Default: "5s".
# Increase if your LLM server is slow to respond.
# timeout = "5s"

# Enable model thinking mode for chat (e.g. qwen3 <think> blocks).
# Unset = don't send (server default), true = enable, false = disable.
# thinking = false

[documents]
# Maximum file size for document imports. Accepts unitized strings or bare
# integers (bytes). Default: 50 MiB.
# max_file_size = "50 MiB"

# How long to keep extracted document cache entries before evicting on startup.
# Accepts "30d", "720h", or bare integers (seconds). Set to "0s" to disable.
# Default: 30d.
# cache_ttl = "30d"

[extraction]
# Model for document extraction. Defaults to llm.model. Extraction works well
# with small, fast models optimized for structured JSON output.
# model = "qwen2.5:7b"

# Timeout for pdftotext. Go duration syntax: "30s", "1m", etc. Default: "30s".
# Increase if you routinely process very large PDFs.
# text_timeout = "30s"

# Maximum pages to OCR for scanned documents. Default: 20.
# max_extract_pages = 20

# Set to false to disable LLM-powered extraction. Text extraction and OCR
# still run independently.
# enabled = true

# Enable model thinking for extraction. Default: false (faster, no <think>).
# thinking = false

[llm]
extra_context = """
My house is a 1920s craftsman bungalow in Portland, OR.
All costs are in USD. Property tax is assessed annually in November.
The HVAC system is a heat pump (Mitsubishi hyper-heat) -- no gas furnace.
"""