Read 2026-10-23
onyx-dot-app/onyx: technical due diligence, read from its repository
What Borehole read in onyx-dot-app/onyx's code and history at 4a750ae: its licence, security advisories, releases, review practice, tests, dependencies and operations. Every finding cites the file and line it rests on. The full report also covers the team and asks the questions the code cannot answer.
| Dimension | Rating |
|---|---|
| Architecture & Codebase — including AI provenance | strong |
| Engineering Process & SDLC Maturity | strong |
| Product & Engineering Maturity | strong |
| Security & Reliability Posture | adequate |
| Observability & Operations | adequate |
| Fit with the Acquirer | not assessable |
What the repository says
- Releases are cut from another branch than the one surveyed
LICENSE: Relicensed: AGPL-3.0 → MIT (2023-05-10)backend/ee/LICENSE: Part of the code is under a separate licence: backend/ee/, web/src/app/ee/, web/src/ee/
Architecture & Codebase — including AI provenance
strongbackend/onyx/db/models.py is 7,805 lines, 72× this codebase's median file
adequateThe median source file here is 107 lines. 79 files exceed 1,070 lines, the largest being 7,805. Size is not a defect on its own — some problems genuinely live in one place — but these are where merge conflicts, review fatigue and single-owner knowledge concentrate, and they are the first thing that slows an inheriting engineer down.
backend/onyx/db/models.py:1 — 7,805 lines (72× median) |
web/src/app/craft/hooks/useBuildSessionStore.ts:1 — 3,772 lines (35× median) |
backend/onyx/server/features/mcp/api.py:1 — 3,598 lines (33× median) |
backend/onyx/auth/users.py:1 — 3,393 lines (31× median) |
backend/onyx/server/manage/llm/api.py:1 — 3,301 lines (30× median) |
Engineering Process & SDLC Maturity
strongTests are retried until they pass
adequateThe suite is configured to retry failing tests (web/playwright.config.ts). A retry turns a flaky test green without saying why it failed, so a green build means the tests passed eventually. Ask how many retries a typical run needs, and which tests need them.
web/playwright.config.ts:18 — test retry |
Product & Engineering Maturity
strongReleases are cut from another branch than the one surveyed
adequateThe newest release, v4.9.4, is not on the branch this survey read; the newest release on it is v4.2.0. The shipped product may differ from what this report describes. Survey the release branch too.
| tag v4.9.4 |
Security & Reliability Posture
adequate21 credential-shaped strings in the tree sit in tests, examples, documentation or local development settings
adequateStrings shaped like credentials appear only where fakes are usual: test fixtures, examples, documentation, placeholders and local stacks. They are almost always fakes. Ask someone to confirm that none of them works anywhere, because a real key in a development file is still a real key. 3 other matchs set aside: expired tokens, licence tokens a vendor verifies with a public key, or placeholder text where a key would go. None grants access to anything.
| 2 cited: AWS access key, where fakes are usual |
| 3 cited: Slack token, where fakes are usual |
1 Dockerfile runs the application as root
adequateNo USER instruction switches away from root, so the process runs with full privileges inside the container. It is a one-line fix and an ordinary expectation, which is why its absence is worth asking about: it usually says more about whether anyone has reviewed the deployment than about this container specifically.
backend/Dockerfile — runs as root |
The build runs third-party code it does not pin
adequate3 steps pipe a download straight into a shell. Tags have been moved to malicious commits in real supply-chain attacks, and a piped script is whatever the server returns that minute. Pinning to a commit is the fix, and it is small.
deployment/docker_compose/install.sh:8 — download piped into a shell |
12 outbound calls can wait forever
adequate12 HTTP calls use a client with no timeout (Python `requests` without `timeout=`, or Go's default client). When a dependency slows down, each one holds a worker until the connection gives up, and enough of them take the service down with the dependency.
backend/ee/onyx/server/oauth/confluence_cloud.py:208 — outbound call without a timeout |
backend/ee/onyx/server/oauth/confluence_cloud.py:300 — outbound call without a timeout |
backend/ee/onyx/server/oauth/google_drive.py:168 — outbound call without a timeout |
4 deserialisers that can run code
adequate4 places read data with a deserialiser that executes what it reads (pickle). That is safe on data the system wrote itself and remote code execution on anything else. Ask where each one's input comes from.
backend/onyx/background/celery/configs/base.py:190 — code-executing deserialiser |
backend/onyx/indexing/chunk_batch_store.py:55 — code-executing deserialiser |
backend/onyx/utils/isolated_runner.py:26 — code-executing deserialiser |
30 locked dependencies have published vulnerabilities, 4 high or critical in what ships
adequate30 of 1,120 locked dependencies match 72 published advisories in OSV, 17 of them rated high or critical. A lockfile says which version is installed, not whether the flawed code is reachable from this product, so each needs a look. Most are fixed by an upgrade, and the advisory names the version. This lockfile pins development tools alongside what ships and does not say which is which, so some of these may never reach a user.
11 places insert HTML from a variable without sanitising it
adequate11 places set HTML from a value that is neither a literal nor passed through a sanitiser. Where the value can come from a user, that is stored cross-site scripting, and it is the first thing a buyer's security review looks for in a frontend.
extensions/chrome/src/utils/error-modal.js:267 — HTML from a variable |
extensions/chrome/src/utils/error-modal.js:329 — HTML from a variable |
web/src/app/app/message/messageComponents/renderers/CustomToolRenderer.tsx:51 — HTML from a variable |
Relicensed: AGPL-3.0 → MIT (2023-05-10)
adequateThe root licence changed 1 time(s). Code taken under earlier terms stays under them; check which terms any fork or customer relied on.
Part of the code is under a separate licence: backend/ee/, web/src/app/ee/, web/src/ee/
adequateA directory named for a paid edition carries its own licence. That code is usually not open source even where the rest is; what the buyer may do with each part differs.
backend/ee/LICENSE — licence: unrecognised |
web/src/app/ee/LICENSE — licence: unrecognised |
web/src/ee/LICENSE — licence: unrecognised |
6 advisories against this project in the last 24 months, 2 high or critical
adequateThe newest was published on 2026-09-04. A project that publishes advisories is telling its users about its fixes, and the count and pace say how much security work it carries. Whether 4 of them are fixed in the code you ship could not be told from this history: its version is only known to be at least 4.2.0. Answer lives in: the release you run, set against each fixed version. GHSA-7f48-vgpj-h95m (moderate) is fixed in the surveyed commit; which release carries it cannot be told from this history, which does not place the release users run. Answer lives in: the version you run. The record is incomplete: NVD: 7 records name onyx with nothing tying them to this repository (CVE-2020-11584, CVE-2023-43784, CVE-2024-0101, …).
| 2026-09-04: CVE-2026-85700 (high): Onyx 4.6.6 fails to properly restrict access to custom tool credentials stored in custom_headers, allowing any authenticated user to read admin-defined API keys |
| 2026-07-20: GHSA-7f48-vgpj-h95m (moderate), fixed in 4.3.0: Curator-scope IDOR: any curator can modify membership of arbitrary user groups via unscoped PATCH /manage/admin/user-group/{id} and /add-users leading to cross- |
| 2026-07-20: GHSA-q62f-rv3h-f822 (critical), fixed in 4.0.0+: Cross-user OAuth-token leak via /api/mcp/servers* for per-user MCP servers |
| 2026-04-29: GHSA-rw6w-hp62-gc8w (moderate), fixed in 3.0.9: IDOR in /chat/stop-chat-session allows any authenticated user to interrupt other users chat sessions |
| 2026-04-29: GHSA-vg3h-35f7-7w6r (moderate), fixed in 3.0.9: IDOR in /chat/file/{file_id} allows any authenticated user to download other users files |
Observability & Operations
adequateSends usage data to telemetry.onyx.app by default
adequatebackend/onyx/utils/fleet_telemetry.py sends to telemetry.onyx.app, and backend/onyx/background/celery/tasks/monitoring/tasks.py leaves reporting on as shipped. Every installation reports to the vendor unless its operator turns this off. For a buyer running it inside their own product, that is an outbound data flow to a third party: ask what is sent, how often, and whether the buyer's customers were told.
backend/onyx/utils/fleet_telemetry.py:35 — reports to telemetry.onyx.app |
backend/onyx/background/celery/tasks/monitoring/tasks.py:324 — the reporting setting is on by default |
4 of 14 deployed images carry no version, or `latest`
adequate4 images in the deployment configuration are named without a version or as `latest`. A redeploy can change the software without a commit, a rollback has nothing to roll back to, and nobody can say from the repository what is running.
deployment/docker_compose/docker-compose.prod.yml:470 — image without a version |
deployment/helm/charts/onyx/Chart.yaml:19 — image without a version |
deployment/helm/charts/onyx/Chart.yaml:21 — image without a version |
Readiness and liveness probes are identical in 2 of 12 workloads
adequateBoth probes call the same endpoint. When a dependency slows down, the liveness probe fails with the readiness probe, so the platform restarts pods that only needed taking out of rotation, and a partial outage becomes a full one. Liveness should ask whether the process is alive; readiness whether it can serve.
deployment/helm/charts/onyx/templates/mcp-server-deployment.yaml — same probe endpoint twice |
deployment/helm/charts/onyx/templates/sandbox-proxy/deployment.yaml — same probe endpoint twice |
Schema migrations run every time a container starts
adequateThe start command runs the database migration and then the server, and the deployment runs more than one replica. Replicas race to migrate, a rollback cannot undo the schema, and a failed migration takes the service down with it. A separate job or release step is the usual fix.
cli/internal/deploy/deployfiles/embedded/docker_compose/docker-compose.prod.yml:19 — migration in the start command |
cli/internal/deploy/deployfiles/embedded/docker_compose/docker-compose.yml:56 — migration in the start command |
Fit with the Acquirer
not assessableNot rated: structurally not assessable from a repository.
Read more, or read your own
The full report on onyx-dot-app/onyx: every dimension, the questions for the room, and the audit log.
Survey a repository: free on public repositories. Private code is read on your own machine by a collector whose source you can read first.