Enjoy our curated collection of examples and solutions. Use these patterns to build your own robust and scalable applications on Corbits.

Four directories, with clear rules about what belongs in each:

The distinction between the last two matters: modules/ answers "does this package mount?", solutions/ answers "is it useful to an agent?".

The @intx/* Interchange packages come from npm, pinned at 0.2.2 in the root package.json catalog — the version the three @corbits/* cores declare as their peer. The cores themselves are not on npm yet, so this repo vendors them as git submodules and resolves them as bun workspace members:

git submodule update --init --recursive
bun install

bun install also builds the three @corbits/* cores (their dist/ is gitignored). Re-run it after checking a submodule out at a new commit.

Then:

bun run typecheck    # every example
bun run modules      # all five modules/ entries, against a real Postgres

The modules/ entries and solutions/inbox-triage need a Postgres — see the modules README for a one-liner that starts one.

The agent-shaped examples take their inference configuration from the environment. This repo ships no default provider, endpoint or model — run one unconfigured and it exits immediately telling you what to export, rather than guessing at something that happens to be installed.

INTX_BASE_URL wins over a vendor key. Model ids are never defaulted — they move faster than this repo does, so INTX_MODEL is the one value you always name yourself.

Start from .env.example, which has a block per provider with none of them active:

cp .env.example .env       # uncomment exactly one block, then
set -a; . ./.env; set +a   # export it into your shell

Or export directly. A local Ollama — no account, no key that anyone checks:

ollama serve                                  # if it is not already running
ollama pull qwen2.5vl:7b                      # or any chat model you like
export INTX_BASE_URL=http://localhost:11434/v1
export INTX_MODEL=qwen2.5vl:7b
export INTX_API_KEY=ollama                    # Ollama ignores it; the field is required

A hosted vendor:

export ANTHROPIC_API_KEY=sk-ant-...
export INTX_MODEL=claude-sonnet-4-6

Anything else that speaks the OpenAI wire format — vLLM, LM Studio, OpenRouter, a gateway of your own — is the same three variables:

export INTX_BASE_URL=https://openrouter.ai/api/v1
export INTX_MODEL=meta-llama/llama-3.3-70b-instruct
export INTX_API_KEY=sk-or-...

There is no vendor branch anywhere in the runtime: provider selects one of the wire-format adapters @intx/inference ships, and the other three fields are configuration handed straight through. See @corbits/example-kit — the whole resolver is one file, and it names no default endpoint.

Then run any agent example:

cd starter/agent-quickstart && bun run start "name three planets"
cd starter/workflow-quickstart && bun run start "the rings of Saturn"
cd solutions/inbox-triage && bun run start

The examples share a host, assertion helpers and the inference resolver, all from @corbits/example-kit. They do not share the mount. Every modules/ entry keeps its own src/app.ts holding the mountX call and the seams that host supplies — the identity resolver, the content store, the triage vocabulary — because those are the decisions an adopter has to make, and hiding them behind a helper would defeat the point.

src/index.ts next to it is the proof: it drives the app in-process and asserts. bun run serve binds the same app to a port.

Each example should be self-contained and include:

• A .gitignore
• A package.json with the license set to MIT
• A README.md with a short description and, if it requires environment variables, a .env.example file and instructions on how to set them up

• Corbits Docs
• Faremeter Docs

If you have any questions or suggestions about the docs, feel free to open a discussion, or submit a PR with your suggestions!

• Start a Discussion with a question, piece of feedback, or idea you want to share with the team.
• Open an Issue if you believe you've encountered a bug that you want to flag for the team.