ktx/README.md
Andrey Avtomonov 494618ab14
feat: add codex llm backend for ktx runtime work (#253)
* feat: add codex sdk runner foundation

* feat: parse codex runtime events

* feat: expose codex runtime mcp tools

* feat: add codex llm runtime

* feat: wire codex llm backend

* test: avoid Array.fromAsync in codex runner test

* docs: document codex llm backend

* fix: tighten codex runtime config ownership

* fix: use codex sdk env and thread options

* fix: parse codex sdk event shapes

* test: add codex backend live smoke

* docs: clarify codex backend isolation

* fix: drive codex loop metrics from mcp events

* fix: enforce codex local step budget

* docs: disclose codex isolation limits

* fix: count all codex agent steps and stream step callbacks live

The agent-loop step budget only counted completed mcp_tool_call items, so
built-in command_execution steps (which the public Codex SDK/CLI surface can
still expose) never decremented the budget, letting ingest/reconciliation run
past stepBudget until Codex stopped on its own. onStepFinish was also replayed
only after the whole stream drained, so live work_unit_step / reconciliation
progress appeared stuck until the Codex process exited.

collectEvents is now the single live step accumulator: it counts every
completed agent-action item via a shared isCompletedAgentStep predicate
(command_execution, mcp_tool_call, file_change, web_search), fires onStepFinish
as each step completes, and enforces the budget on that broader count. A
no-tool turn still counts as one step. toolFailures stays MCP-specific, since a
non-zero command exit is normal agent exploration, not a loop failure.

* test: align ingest llm-guard assertions with codex backend

The skip-llm ingest guard message now lists codex as a valid backend and
mentions a Claude Code/Codex session plus a codex setup hint, but this slow
suite test still asserted the pre-codex wording. Update it to match the
production message (already covered by the local-bundle-runtime unit test) and
add the codex setup-line assertion.

* fix: treat codex error:null tool calls as success

The Codex SDK serializes error: null on successful mcp_tool_call items, so
the failure check (item.error !== undefined) flagged every successful tool
call as failed with the empty-payload default "Codex turn failed". This
killed every ingest work unit under the codex backend before it could
produce a patch.

Key on status === 'failed' (authoritative, always set) and only treat a
populated error object as a failure. Add a regression test built from a
verbatim real-SDK event capture.

* fix: default codex backend to gpt-5.5 and report real probe errors

The previous default gpt-5.3-codex is an API-key-only model that the OpenAI
API rejects under ChatGPT-account (subscription) auth, so codex status/setup
failed with a misleading "authentication is not usable" message even though
auth was fine.

- Default codex model is now gpt-5.5 (works on both subscription and API-key
  auth); the curated setup picker offers gpt-5.5 / gpt-5.4 / gpt-5.4-mini and
  keeps free-form entry for account-specific ids (e.g. gpt-5.3-codex-spark).
- runCodexAuthProbe now distinguishes "model not available" from an auth
  failure and surfaces the real API error: collectEvents retains stream
  events when the SDK throws on a non-zero exit, and the API error JSON
  envelope is unwrapped to its human-readable message.
- The Codex isolation warning now renders inside the clack setup frame.
- Docs updated to gpt-5.5 with a note that *-codex ids require API-key auth.

* fix: require llm.models.default in status and match codex probe remediation

Status reported a project ready when a non-none LLM backend was configured
without llm.models.default, but the runtime (resolveModelSlots) hard-requires
it, so ingest/scan/memory threw after `ktx status` said the project was usable.
buildLlmStatus now fails for any non-none backend missing models.default and no
longer invents a fallback model for claude-code/codex.

Codex probe failures now carry a category-matched fix: a model-access failure
steers the user at llm.models.default instead of the auth/install remediation.
runCodexAuthProbe returns the fix and status consumes it; the message stays
self-sufficient so setup output is unchanged.

Docs: README now lists the codex backend and local Codex auth; ktx-setup.mdx
states --llm-model only accepts codex/default or gpt-*/codex-* ids.

Repaired four doctor fixtures that configured a backend without models.default
(the now-correctly-blocked config) and added coverage for the new behavior.
2026-06-02 13:57:11 +02:00

266 lines
10 KiB
Markdown

<h1 align="center">
<img src="assets/ktx-lockup.svg" alt="ktx" width="500" />
</h1>
<h1 align="center">
The context layer for data agents
</h1>
<p align="center">
<a href="https://www.npmjs.com/package/@kaelio/ktx"><img src="https://img.shields.io/npm/v/@kaelio/ktx?style=flat-square&color=f97316" alt="npm version" /></a>
<a href="https://codecov.io/gh/Kaelio/ktx"><img src="https://codecov.io/gh/Kaelio/ktx/graph/badge.svg?branch=main" alt="Codecov" /></a>
<a href="https://github.com/Kaelio/ktx/actions/workflows/ci.yml?query=branch%3Amain"><img src="https://img.shields.io/github/actions/workflow/status/Kaelio/ktx/ci.yml?branch=main&label=tests&style=flat-square" alt="Tests" /></a>
<a href="https://docs.kaelio.com/ktx/docs/"><img src="https://img.shields.io/badge/docs-ktx-22c55e?style=flat-square" alt="Documentation" /></a>
<a href="https://join.slack.com/t/ktxcommunity/shared_invite/zt-3y9b44m1x-LVyNNJD5nwaZHq4XS29LMQ"><img src="https://img.shields.io/badge/slack-join%20community-4A154B?style=flat-square&logo=slack&logoColor=white" alt="Join the ktx Slack community" /></a>
<a href="https://github.com/Kaelio/ktx/blob/main/LICENSE"><img src="https://img.shields.io/badge/license-Apache%202.0-blue?style=flat-square" alt="License" /></a>
<a href="https://www.ycombinator.com/companies?batch=P25"><img src="https://img.shields.io/badge/Y%20Combinator-P25-orange?style=flat-square" alt="Y Combinator P25" /></a>
</p>
<p align="center">
<a href="https://docs.kaelio.com/ktx/docs/getting-started/quickstart"><b>Quickstart</b></a> ·
<a href="https://docs.kaelio.com/ktx/docs/cli-reference/ktx"><b>CLI Reference</b></a> ·
<a href="https://docs.kaelio.com/ktx/docs/ai-resources/agent-quickstart"><b>Agent Setup</b></a> ·
<a href="https://join.slack.com/t/ktxcommunity/shared_invite/zt-3y9b44m1x-LVyNNJD5nwaZHq4XS29LMQ"><b>Slack</b></a>
</p>
---
**ktx** is a self-improving context layer that teaches agents how to query your
warehouse accurately - from approved metric definitions, joinable columns, and
business knowledge it builds and maintains for you.
> [!NOTE]
> Run **ktx** with your own LLM API keys or a local agent sign-in — a
> **Claude Pro/Max** subscription through Claude Code, or your local Codex
> authentication. No extra usage billing from **ktx**.
<p align="center">
<a href="https://youtu.be/5V4TuzYVlrA">
<img src="assets/launch-video-thumb.png" alt="Watch the ktx launch video (1:56)" width="820" />
</a>
</p>
<p align="center">
<img src="docs-site/public/images/ingestion-flow.png" alt="Ingestion: ktx ingests databases, BI tools, modeling code, and docs through its context engine (source connectors, context builder, reconciliation, validation) into wiki Markdown and semantic-layer YAML" width="900" />
</p>
<p align="center">
<img src="docs-site/public/images/mcp-runtime-flow.png" alt="Serving: an agent queries ktx through MCP, which searches the wiki and semantic layer, returns approved metrics, and compiles them into read-only SQL run against the warehouse" width="900" />
</p>
## Why ktx
General-purpose agents struggle on data tasks. They re-explore your warehouse
on every question, invent their own metric logic, and return numbers that
don't match approved definitions.
Traditional semantic layers don't fix this. They demand constant manual
upkeep and don't absorb the rest of your company's knowledge.
**ktx** does both, automatically:
- **Learns from company knowledge.** Ingests wiki content, organizes it,
removes duplicates, and flags contradictions for human review.
- **Maps the data stack.** Samples tables, captures metadata and usage
patterns, detects joinable columns, and annotates sources so agents write
better queries.
- **Builds a semantic layer.** Combines raw tables and high-level metrics
through a join graph that automatically resolves chasm and fan traps, so
agents fetch metrics declaratively instead of rewriting canonical SQL each
time.
- **Serves agents at execution.** Exposes CLI and MCP tools with combined
full-text and semantic search across wiki and semantic-layer entities.
## How ktx compares
| | General-purpose agent | Traditional semantic layer | **ktx** |
| --- | :---: | :---: | :---: |
| Builds warehouse context automatically | — | — | ✓ |
| Detects joinable columns + resolves fan/chasm traps | — | Manual | ✓ |
| Approved, reusable metric definitions | — | ✓ | ✓ |
| Absorbs wiki / Notion / team knowledge | — | — | ✓ |
| Flags contradictions across sources | — | — | ✓ |
| Ships CLI + MCP for agent execution | Partial | — | ✓ |
| Read-only by design | n/a | n/a | ✓ |
## Who is ktx for
**Use ktx if you:**
- Want agents like Claude Code, Codex, Cursor, or OpenCode to query your
warehouse with approved metric definitions
- Have business knowledge scattered across dbt, Looker, Metabase, Notion, and
team wikis
- Need agents to reuse canonical SQL instead of inventing it on every prompt
**Skip ktx if you:**
- You don't have a SQL warehouse - **ktx** sits on top of one
- You only need one ad-hoc query - `psql` or a notebook will do
Works with PostgreSQL, Snowflake, BigQuery, ClickHouse, MySQL, SQL Server, and
SQLite. Integrates with dbt, MetricFlow, LookML, Looker, Metabase, and Notion.
## Quick Start
```bash
npm install -g @kaelio/ktx
ktx setup
ktx status
```
`ktx setup` creates or resumes a local **ktx** project, configures providers
and connections, builds context, and installs agent integration.
Example `ktx status` after setup:
```text
ktx project: /home/user/analytics
Project ready: yes
LLM ready: yes (claude-sonnet-4-6)
Embeddings ready: yes (text-embedding-3-small)
Databases configured: yes (warehouse)
Context sources configured: yes (dbt_main)
ktx context built: yes
Agent integration ready: yes (codex:project)
```
> [!TIP]
> Already using an agent? Ask Claude Code, Codex, Cursor, or OpenCode from
> your project directory:
>
> ```text
> Run npx skills add Kaelio/ktx --skill ktx and use the ktx skill to install
> and configure ktx in this project.
> ```
> [!IMPORTANT]
> If `ktx status` prints `ktx mcp start --project-dir ...`, run it before
> opening your agent client.
## First commands
| Command | Purpose |
| --- | --- |
| `ktx setup` | Create, resume, or update a **ktx** project |
| `ktx status` | Check project readiness |
| `ktx ingest` | Build context for every configured connection |
| `ktx sl "revenue"` | Search semantic sources |
| `ktx wiki "refund policy"` | Search local wiki pages |
| `ktx mcp start` | Start the MCP server for agent clients |
See the [CLI Reference](https://docs.kaelio.com/ktx/docs/cli-reference/ktx)
for every command, flag, and option.
## Project Layout
```text
my-project/
├── ktx.yaml # Project configuration
├── semantic-layer/<connection-id>/ # YAML semantic sources
├── wiki/global/ # Shared business context
├── wiki/user/<user-id>/ # User-scoped notes
├── raw-sources/<connection-id>/ # Ingest artifacts and reports
└── .ktx/ # Local state and secrets, git-ignored
```
Commit `ktx.yaml`, `semantic-layer/`, and `wiki/`. Keep `.ktx/` local.
Project resolution defaults to `KTX_PROJECT_DIR`, then the nearest `ktx.yaml`,
then the current directory. Pass `--project-dir <path>` when scripting.
## FAQ
- **Does ktx send my schema or query results to a hosted service?**
No. **ktx** runs locally. The only data leaving your machine is what you
send to the LLM provider you configured.
- **Which LLM backends are supported?**
Anthropic API, Google Vertex AI, AI Gateway, the local Claude Code session
through the Claude Agent SDK, and your local Codex authentication through the
Codex SDK. See
[LLM configuration](https://docs.kaelio.com/ktx/docs/guides/llm-configuration).
- **How is ktx different from a dbt or MetricFlow semantic layer?**
**ktx** *ingests* those layers and combines them with raw-table
introspection and wiki content. Agents get one searchable surface instead
of three disconnected ones - and **ktx** flags contradictions across
sources.
- **Does ktx need a running server?**
There is no hosted service. The local MCP daemon runs on demand via
`ktx mcp start` when an agent client needs it.
- **Is my warehouse safe?**
Yes. Connections are read-only - **ktx** never writes to your database.
## Docs
- [Quickstart](https://docs.kaelio.com/ktx/docs/getting-started/quickstart)
- [The Context Layer](https://docs.kaelio.com/ktx/docs/concepts/the-context-layer)
- [Building Context](https://docs.kaelio.com/ktx/docs/guides/building-context)
- [CLI Reference](https://docs.kaelio.com/ktx/docs/cli-reference/ktx)
- [Agent Quickstart](https://docs.kaelio.com/ktx/docs/ai-resources/agent-quickstart)
- [Community & Support](https://docs.kaelio.com/ktx/docs/community/support)
## Community
- **[Slack](https://join.slack.com/t/ktxcommunity/shared_invite/zt-3y9b44m1x-LVyNNJD5nwaZHq4XS29LMQ)** — ask questions, share what you're building, and chat with maintainers.
- **[GitHub Issues](https://github.com/Kaelio/ktx/issues)** — report bugs and request features.
- **[Contributing](https://docs.kaelio.com/ktx/docs/community/contributing)** — set up the repo, run tests, and open a PR.
## Development
```bash
git clone https://github.com/kaelio/ktx.git
cd ktx
pnpm install
uv sync --all-groups
pnpm run build
pnpm run check
```
**ktx** is a pnpm + uv workspace:
| Path | Purpose |
| --- | --- |
| `packages/cli` | TypeScript CLI and published npm package source |
| `packages/cli/src/context` | Core context engine |
| `packages/cli/src/llm` | LLM and embedding providers |
| `packages/cli/src/connectors` | Database scan connectors |
| `python/ktx-sl` | Semantic-layer query planning |
| `python/ktx-daemon` | Portable compute service |
Local development CLI:
```bash
pnpm run setup:dev
pnpm run link:dev
ktx-dev --help
```
Useful checks:
```bash
pnpm run type-check
pnpm run test
pnpm run dead-code
uv run pytest -q
```
## Telemetry
**ktx** collects anonymous usage telemetry from interactive CLI runs to
improve setup, command reliability, and data-agent workflows. No file paths,
hostnames, SQL, schema names, error messages, or argv are recorded. See
[Telemetry](https://docs.kaelio.com/ktx/docs/community/telemetry) for the
event catalog and opt-out options.
## License
**ktx** is licensed under the Apache License, Version 2.0. See `LICENSE`.
## Star History
<p align="center">
<a href="https://star-history.com/#Kaelio/ktx&Date">
<img src="assets/star-history.svg" alt="ktx Star History Chart" width="700" />
</a>
</p>