Walkthrough

Local setup. Saves an encrypted settings file and asks only where the graph lives (default ~/.oynix/local), then offers to wire your agent. No provider prompt. You start key-free with a structural graph (files, functions, call + import edges) and set the LLM + embedder later in the Models UI. Add a key and re-index (or the connectors page's enrich) to backfill docstrings + embeddings onto the same graph. Nothing is thrown away.

✓ Wrote ~/.oynix/config.enc    Oynix is ready. Next: oynix connect

A scripted/CI run can still pass --llm-provider/--llm-key/--llm-model/--llm-url (+ --embeddings-*) to bake a key in non-interactively. Interactive local never asks. That's the Models UI's job.

Starts the engine if needed and prints a URL to the connectors page, then opens it in your browser. Add git repos (they index there, with progress) and doc sources like Jira, Slack, and Google Drive. This is the one surface for adding code and context. Here is the page you open:

Oynix

Connectors

tuneModelscableConnectorshubGraph

Oynix

search

Show

apps All 6check_circle Connected 0radio_button_unchecked Not connected 6

Connect the tools your team already uses. Their tickets, docs and threads become part of the graph and link to the code they're about.

GitHubradio_button_unchecked Not connected

Google Driveradio_button_unchecked Not connected

Confluenceradio_button_unchecked Not connected

Jiraradio_button_unchecked Not connected

Files & Docsradio_button_unchecked Not connected

computerThis machineradio_button_unchecked Not connected

cableSelect a connector to see its details, sync status and actions.

The one launcher for every web surface. Every page shares one top nav, so once any is open you can tab across to the rest. The Models page is where you set your language model and embedder, live with no restart, so you can start key-free and add a provider whenever. Here is that page:

Oynix

Models

tuneModelscableConnectorshubGraph

Oynix

Models

edit_noteLanguage model Not set yetscatter_plotEmbedder Not set yet

edit_noteLanguage model

Changeable anytime

What it does

Writes the plain-English summaries (docstrings) for your code, and powers the answers in Ask. It turns raw code into something you can ask questions about.

The better this model, the better every explanation, root cause, and diagram Oynix produces.

Without it: Indexing still builds the structural graph (files, functions, call and import edges), but skips the summaries, so Ask has less to work with.

Status Not set yet

ProviderBring your own key

EndpointAny OpenAI-compatible URL

Start key-free. Add a key here anytime. Live, no restart.Set a key

Wire a coding agent to your local engine over MCP. Writes the client's config so your agent answers from your own indexed memory. init already offered to do this; run it any time to add another agent.

✓ Registered with Claude Code. Restart it to load the server.

See the knowledge graph. Running oynix graph opens this view:

The embedded engine lifecycle. It auto-starts when you connect or open the graph, so you rarely type start yourself. Stop and status are the escape hatches.

One health summary: the embedded engine, a real store probe, your LLM + embedder providers, and capability coverage.

Oynix Status

engine       ✓ running     (embedded)
store        ✓ ready       (~/.oynix/local)
providers    LLM ✓ qwen-plus    embedder ✓ 1024-d

Capability coverage
  ✓ Code search   ✓ Docs ingest   ○ Vision (add a vision provider)

Runs checks and prints pass/fail with a fix hint for each failure.

The encrypted settings file.