AI & Edge
daedalus can package AI and edge workloads into a single self-contained binary. This includes local models via Ollama and MCP tools for agents.
On-device AI with Ollama
Package applications that run local models via Ollama or llama.cpp. The binary embeds the model files and the Ollama runtime. On first run, the model executes on the target machine with no cloud dependency.
Use case: offline code analysis
1. Write a Python/Node app that uses the daedalus-core hidden-dependencies
analyzer to scan source code for subprocess/dlopen calls.
2. The analyzer runs a local model via Ollama for LLM-powered classification.
3. Package with daedalus build . -o app.daedalus.
4. The .daedalus binary embeds Ollama + the model.
5. On first run, Ollama starts and serves the analyzer on
http://127.0.0.1:PORT.
6. The app performs offline code analysis — no internet, no cloud, no cost.
# Build the app with Ollama embedded
daedalus build ./my-code-assistant -o code-assistant.daedalus
# First run: Ollama auto-starts, model loads
./code-assistant.daedalus
# Then analyze any codebase offline
code-assistant.daedalus scan /path/to/codebase
Ollama detection
daedalus detects Ollama-based apps automatically:
ollamainpackage.jsonscripts or dependenciesModelfileormodels/directory
The binary embeds the Ollama binary and model files. At runtime, it starts the Ollama server and serves the app on the configured port.
daedalus build ./my-ollama-app -o my-ollama-app.daedalus
MCP tools for agents
daedalus can embed Model Context Protocol (MCP) tools as standalone binaries or scripts. They are exposed over stdin/stdout JSON-RPC.
daedalus build ./my-agent -o my-agent.daedalus \
--mcp-tools ./tools/