Skip to content

Releases

Release v0.9.0

Choose a tag to compare

@homer6 homer6 released this 16 Nov 10:09
a881cc3

2026-11-16. The open-source release: lake is public under the MIT licence, the security fixes from the pre-launch review, and lake running on liquid v0.3.0. Tracking issue: lake#1344.

Built against liquid v0.3.0 (66b5bb9) and zetta v0.5.0 (e11c02f5), the commits lake-platform pins. Not run on the dev environment before the cut (Steve's decision): the image waits for the operator changes under Before it rolls.

What it includes

Open source

  • MIT LICENSE and COPYRIGHT (OramaSearch, Inc.), an SPDX-License-Identifier: MIT line in every source file lake wrote (tools/spdx-headers.py), THIRD_PARTY_LICENSES and an SBOM (sbom.spdx.json, sbom.cdx.json), CONTRIBUTING.md with the CLA, and CODE_OF_CONDUCT.md (#1321).
  • SECURITY.md: report privately, to security@orama.com or through GitHub's private vulnerability reporting, which is on (#1290).
  • The README is the public front door, linking Orama Lake (#1356). Internal material is out of the tree: runbooks, pricing documents, people, hosts, clusters, private links, review finding ids and quoted decisions (#1317, #1320 and a final sweep); tools/release/public-scan.sh checks it, and the gate page is docs/open-source/ (#1335). Submodules are fetched over HTTPS (#1359).
  • Neutral defaults: no Orama sender, support address, legal links or order sheet unless configured (LAKE_RESEND_FROM, LAKE_SUPPORT_EMAIL, LAKE_TERMS_URL, LAKE_PRIVACY_URL, LAKE_ORDER_SHEET) (#1320). The schema seeds no model prices. The hosted CI workflows are commented out (#1336).

Security fixes (each with its test; the findings are closed)

  • Sign-in and sessions: cross-site sign-in posts refused; __Host- session cookies behind TLS; a Clerk token issued to another product is not exchanged; a new Clerk subject is never linked to an existing account by email; a rotated-away TAAP key stops signing in (#1279, #1282, #1286, #1296, #1300).
  • Agents: a person's service launch is called with its own token, never the Lake's (#1276); task, launch and schedule credentials act only in their own account and workspace, through the workspace a call names, its route and the object it names, a public repository of another workspace included (#1286, #1292, #1296, #1356); a delegated turn never holds more than the asking agent; task credentials can be passed by reference and bus credentials bound to one workspace (#1304); a dataset-confined launch's query carries its work area to the engine (#1306).
  • Repositories and drives: visibility needs admin (#1337); Drive history authorizes the node at every revision; Drive, methodology, model and store writes follow branch rules (#1287, #1288); repository reads honour restricted folders; share-link expiry is a real RFC 3339 UTC time (#1339).
  • Platform: custom domains are claimed with a DNS TXT proof before the operator routes them, and a tenant cannot claim one of the Lake's own names (#1284, #1295); a service cannot be routed to another customer's workspace (#1302); an app binds only its own workspace's services (#1289); app token rotation is a generation the operator acknowledges (#1304).
  • PKI, MCP, registry, secrets, orders, billing: ACME needs an account binding and a real HTTP-01 check (#1283); MCP tool access and IPv6 address checks (#1277, #1279); OCI uploads continue only in their repository; a sealed delivery checks the mount's workspace; order forms mail an address only a confirmation until it is confirmed; unauthenticated payment notices are counted, not each audited; a deleted workspace's unbilled usage is kept for its bill (#1278, #1280, #1296); unknown HTTP methods are counted as other (#1340); signed links use a derived key; the SQL guard parses like the engine and the GitHub import reads only the caller's staging folder (#1281).

Liquid v0.3.0

  • One sync server per repository, so liquid's staged uploads meet their authorized push (#1351).
  • Agent session keys meet liquid's delegation/1 rules: mint_agent_session_key takes a required repository_id, Lake is issuer and dispatcher, every key expires within 15 minutes, and never-expiring keys are refused (#1358).
  • A pod seals its liquid identity with a passphrase file (LIQUID_IDENTITY_PASSPHRASE_FILE), and lake stops at start, naming the variables, when it has none (#1355).

Other

  • server_info says which liquid clients this Lake supports (#1330); a change, review or comment shows who signed it by name (#1331); a key is called an API token, with a liquid and liquify preset (#1333); TiDB and Cassandra credentials name a database or keyspace (#1305).
  • Fixes found by the release suite (#1345, #1347–#1350, #1357): dev builds again; a freed Content-Security-Policy string on every web app page; a dangling view that confined every dataset work area to nothing; the registry dropping methodology run and deployment user names; the order email claim statement that never parsed on PostgreSQL; contract, docs and fixture updates.

Migrations

0106_domain_claims, 0107_order_email_confirmation, 0108_retained_usage, 0109_principal_ids, 0110_taap_chain_heads, 0111_service_workspace_routing, 0112_app_token_generation, 0113_agent_session_repositories. The newest guard is agent_session_repositories_guard.

Before it rolls

  • The operator vendors the schema and selects agent_session_repositories_guard.
  • The operator gives each Lake a liquid identity passphrase (orama-operator#1215); without it a new pod does not start.
  • The operator renders LAKE_RESEND_FROM, LAKE_SUPPORT_EMAIL, LAKE_TERMS_URL, LAKE_PRIVACY_URL and LAKE_ORDER_SHEET on Orama's Lakes (orama-operator#1191); rolls together with orama-operator#1157 (sealed deliveries), #1164 (service connections, operator first), #1160 (app bindings) and #1150 (launch service tokens).
  • LAKE_GIT_IMPORT_ROOT turns GitHub import on; unset, it answers not_supported.
  • After the roll: everyone signs in once more (__Host- cookies); agent session keys and signed links made before the roll stop verifying; OCI uploads in flight restart.

Not included

  • #1293 (the engine enforces workspace_sql's read; waits for zetta#1507), #1298 (on-demand service connections; its Clerk rule conflicts with #1296), #1323's model-price docs, #1338 (SCIM selectors, migration 0121) and #1342 (SSO pool binding): open for review.

Known issues

  • Security issues still open are listed in the repository's issues; they are not described here.
  • lake#1353: WorkspaceStoreHandoff.ASearchRebuildDoesNotHoldOffTheRelease crashes intermittently at shutdown (a Zetta object send calls a destroyed server's observer).
  • lake#1334: the image is built on Ubuntu, which bundles GPL and LGPL components.
  • lake#1341: an unknown HTTP method on a routed path closes the connection without an answer.

Verification

  • Full native suite on dev at 565633b0 (macOS, Homebrew Clang 22, PostgreSQL): 1758 run, 1 failed (lake#1353), 28 skipped. Six of the skips were WorkspaceStoreZetta.*, because the Zetta binary used was older than the v0.5.0 pin; the others need an outside service (a Python Hugging Face client, provider credentials) or are Linux-only. The first run on dev after the merges had 87 failures; every one is fixed in the pull requests above.
  • After the cut, against Zetta v0.5.0 (e11c02f5) built from the pin and liquid v0.3.0 (lake#1364): WorkspaceStoreZetta.* 6 of 6 passed, none skipped; WorkspaceStore* 99 of 99.
  • After it, only the changes since were tested where they reach: the cleanup sweep (comments and docs) with the docs, markdown and frontend tests, and #1359's three submodule URLs fetched anonymously over HTTPS.
  • Not run: the Linux lane, the image build and the dev environment.

Release v0.8.1

Choose a tag to compare

@homer6 homer6 released this 27 Oct 18:51
2e6024c

v0.8.1

2026-10-27. A patch release: the fixes from Steve's customer-insights demo on us.lake.live
(lake#1260).

What it includes

  • A file's text is its exact UTF-8 (lake#1238,
    #1265). repository_file (REST and MCP), resolve_model_file
    and resolve_dataset_file answer contents as the file's exact text with binary: false; a file that is not
    UTF-8 has no contents and binary: true. Before, every non-ASCII byte came back as literal \xNN text, and an
    agent that committed what it read wrote the escapes into the file. The drive preview, diff_drive_versions,
    /file/raw and /file/download, methodology sources and the model card use the same rule
    (repositories). Clients: contents is absent for a binary file.
  • Clustering never groups by a row number, an identifier or contact details
    (lake#1242, #1264).
    Columns named like a row number or id are identifiers; clustering also leaves out row numbers recognised by their
    values, free text, names and contact details, and turns a date into its month and its days before the latest.
    With nothing left it refuses, saying why for each column. The demo's "low Index / high Index" segments came from
    this.
  • Predict on a clustering model (lake#1243,
    #1267). The Model page shows a clustering model's segments and
    tables, and Try it becomes Place a Row in a Segment. Against this release's Zetta pin the engine cannot place a
    new row yet, and predict, score and explain say so in plain words (not_supported) instead of a schema
    error; zetta #1499 adds it, in a later Zetta release.
  • Change App reloads onto the new version (lake#1259,
    #1262). The split screen's frame reloads again over the minute
    after the App reads Ready, because the operator reported Ready while the old pod still served (fixed in
    orama-operator #1050). A confirmed proposal now says
    so. New browser check tools/e2e/change-app.mjs.
  • End-to-end fixes: T3 ranks only at-risk tables that have account and risk
    (lake#1258, #1263);
    T12 studies a dataset keyed by a column the study engine recognises
    (agent-experiments#32,
    #1261).
  • The test runner's stale-database sweep no longer breaks a test's connection
    (lake#1254, #1266).

Released with it: agent-application v0.2.7 (files read exactly, agent-application#63; filters applied everywhere,
agent-application#64), app-runtime v0.1.3 (Streamlit messages over 1 MiB, app-runtime#26: the demo App's filter),
and orama-operator v0.10.6 (Ready only when the rollout is complete).

Not included

  • The Zetta release with zetta#1499 (placing a new row in a segment).
  • zetta#1498: a study cannot name its entity column.

Known issues

As in v0.8.0, less those fixed above.

Verification

Each pull request ran lake's full native suite on a fresh local database with a regression test that fails on the old
code: #1265 1643 passed of 1672 (28 skipped need a live Zetta or outside service), #1266 1642 of 1671, #1264 1674 of
1674; #1267 its new test and all 52 Workspaces tests (its full run hit a broken local test template, since removed);
the frontend 469–471 on each.

The release (tag v0.8.1, 2e6024cf): lake's full native suite on a fresh local database: 1678 run, 0 failed,
28 skipped (they need a live Zetta or another outside service), 1 disabled (SearchBenchmark). The image
oramasearch/lake:main-2e6024cf = v0.8.1, sha256:fbb7b502b516ac15ecc1543a3aa21db6d2bc5a52636f9267d9c21517b8c3869c,
built from the tag on the cluster BuildKit, private. Not run: the Linux lane on the release (the affected tests of
#1265 ran there, 14/14).

Where it runs

dev and testing (2026-10-27); staging after the e2e gate on testing. Not production.

Release v0.8.0

Choose a tag to compare

@homer6 homer6 released this 27 Oct 04:58
080d9cc

v0.8.0

2026-10-26. A minor release: agents work on the latest GPT again, and Apps can be changed from the console.

What it includes

  • Agents on gpt-6.1-sol (lake#1245,
    #1253). OpenAI refuses function tools together with
    reasoning_effort on /v1/chat/completions for the gpt-6 models, so every agent turn with tools failed on
    v0.7.2. The model gateway now sends such a request (gpt-6.1-sol, gpt-6.1-terra, gpt-6-astra, gpt-6-luna, with
    tools) to /v1/responses and translates the answer back to a chat completion: tool calls, tool results,
    finish_reason and usage, which is billed exactly as before. gpt-5.5, and gpt-6 calls without tools, stay on
    Chat Completions. The default agent model is gpt-6.1-sol again (model gateway).
  • Edit a repository file (lake#1239,
    #1246): Edit on a text file's preview for people who may
    write to the repository, commit with a message, and Commit and redeploy for a file inside an App's source.
    A file changed on the branch since it was opened is a conflict (expected_blob on commit_repository_files),
    never overwritten. Files are read from content_base64, so non-ASCII bytes stay exact.
  • A conversation that changes an App (lake#1240,
    #1247): a multi-turn thread per App, kept in the workspace;
    Confirm commits and redeploys, and the thread goes on. New capability get_app_conversation; the agent
    receives the earlier turns as history (agent-application v0.2.4 reads them).
  • Change App split screen (lake#1241,
    #1252): the conversation beside the running App, which
    reloads once a confirmed change is deployed and Ready.
  • Everything in v0.7.3 except its gpt-5.5 default, including the version header fix
    (#1251).

Not included

  • Streaming a gpt-6 request that carries tools is answered from the Responses stream; no library agent streams
    with tools today.
  • Any origin can still frame an App: restricting frame-ancestors to the Lake's own origin is an operator
    change, not in this release.

Known issues

  • lake#1254: four SharedLinkScenario tests (drive share
    links) fail on macOS. Not related to this release's changes; under investigation.
  • lake#1242, lake#1243,
    lake#1238 remain open.

Verification

Proof Result
Live round trip against OpenAI on gpt-6.1-sol through the translation Untranslated request: HTTP 400 (the production error). Translated: function_call (80 in / 19 out tokens), then the final answer (126 in / 16 out); streamed call 200; gpt-5.5 with tools stays on Chat Completions, 200
Full native ctest, fresh database (on #1253's branch, which carries every change here) 1665 of 1669 passed; 4 failed (lake#1254); Zetta-backed tests skipped
Frontend suites Run on each feature PR (#1246, #1247, #1252); not re-run on the release branch
Linux lane Not run for this release (Steve: ship without waiting)
Image Built from the tag on the cluster BuildKit

Release v0.7.3

Choose a tag to compare

@homer6 homer6 released this 27 Oct 03:08
25c5c9a

lake v0.7.3

2026-10-26. A hotfix on top of v0.7.2, for a production outage. Image oramasearch/lake:v0.7.3 (private).

What it changes

  • Agents default to gpt-5.5 again (#1248, issue
    #1245).
    • v0.7.2 made gpt-6.1-sol the default agent model.
    • OpenAI refuses a /v1/chat/completions request to the gpt-6.x models that carries function tools with
      reasoning_effort, and the library agents send exactly that. So every agent turn with tools failed on every
      environment: Ask Orama, the Project, Application and Experiments agents, and Quickstart's Ask Orama step.
    • default_agent_model is gpt-5.5. The console labels it as the default.
    • The gpt-6.x models stay choosable per workspace, and the docs say why they are not the default.

Not included

  • The gateway sending a gpt-6.x Chat Completions call that carries tools to /v1/responses. That comes in v0.8.0
    (lake#1244), and makes gpt-6.1-sol the default again.

Known issues

  • A workspace that chooses a gpt-6.x model gets the same refusal on its agents' tool calls until v0.8.0.
  • As in v0.7.2.

Verification

Proof Result
Model gateway, model completion and workspace-setting tests (native) passed; the Postgres variants run in the full suite
Full native suite running at the tag; counts added to this page after the roll
Frontend agent-models tests 5 passed
Found by the v0.7.2 e2e gate on testing: T1–T5, T7, T11, T12 and Quickstart's Ask Orama step
Image built from the tag on the cluster BuildKit

Release v0.7.2

Choose a tag to compare

@homer6 homer6 released this 27 Oct 01:27
ef4a519

Released: 2026-10-26 · Branch: main · Previous: v0.7.1 (2026-10-25) ·
Tracking: #1234.

A patch:

  • An App's workspace.sql(...) works, at the access its binding grants: read-only always, and at most
    5,000 rows a query.
  • Auto-fix sees an App whose code keeps failing while its pod looks healthy, and Restart records what it
    restarted so the operator can replace the pods.
  • People choose the model their workspace's agents use. The gateway serves and prices the gpt-6.x models
    (default gpt-6.1-sol), and a prompt over 272K input tokens is billed at what it costs.

Two migrations (0104_app_restarts, 0105_agent_model_prices), applied by lake migrate from the image. No new
environment variables (Before it rolls).

Fixes

PR Change
#1231 sql on the App gateway runs as workspace_sql, at the binding's access level (#1229, also #1212). sql on /workspaces/<customer>/zetta/v1/ (and under /w/<workspace>/) is no longer unknown_capability: GET takes the input from the query string (what app-runtime sends), POST a JSON body. The App's read binding, the read-only check, safe mode and the audit apply as for people and agents. Writes, DDL, CREATE USER and two statements answer 400 invalid_input and never reach the workspace. A statement without a LIMIT gets LIMIT 5001; more than 5,000 rows, or a LIMIT over 5,000, answers the new error code too_many_rows (422). An answer is never cut short without saying so. A direct workspace_sql call is uncapped. workspace_sql accepts the engine's intent and passes it on (../api/workspaces.md, ../api/web-apps.md, ../api/errors.md)
#1232 Auto-fix sees a healthy App whose script keeps failing, and Restart is recorded (#1228). For a running, Ready App with auto-fix on, the repair pass reads its newest 500 log lines from the last five minutes through app_logs; two or more Streamlit or FastAPI tracebacks record an app_errors notification (the exception line as its reason, the traceback as detail) and start auto-fix's one attempt for that generation. Apps with auto-fix off are never read. Metric lake_repair_log_reads_total. Restart now sets apps.restart_generation (migration 0104_app_restarts), and get_app answers it. The pods are replaced only once orama-operator#1033 is released and rolled (it writes the number on the pod template); until then Restart is recorded but no pod is replaced (../api/notifications.md, ../control-plane/web-apps.md)

Features

PR Change
#1233 People choose the agents' model per workspace; the gateway serves and prices gpt-6.x and bills long prompts at cost (#1230). The gateway serves gpt-6.1-sol, gpt-6.1-terra, gpt-6-astra and gpt-6-luna beside the existing models. A workspace setting agent_model (get_workspace_settings for members, set_workspace_settings for owners and admins), kept as a sealed lake_workspace_settings record in the workspace Zetta (nothing in Postgres), names the model every agent turn's chat call is sent with; unset means gpt-6.1-sol. A chosen model that is no longer served is refused, never swapped. The console's Workspaces → Settings has a picker with prices. Migration 0105_agent_model_prices seeds the rates at OpenAI's standard price plus 30% (read 2026-10-26), with long-prompt classes <model>@long/input and <model>@long/output (also for gpt-5.5): more than 272,000 input tokens is recorded under them. service_tier is pinned to default; any other value is refused. With billing on, a model is served only while its rates, long ones included, are published (../api/model-gateway.md, ../api/billing.md, ../api/workspaces.md)

Docs

PR Change
#1222 The v0.7.1 image and its digest on its release page
#1223 A runbook for the human steps that give app hosts an edge certificate (../deployment/edge-certificates.md, #798); a human action, open

Before it rolls

  • Schema: two migrations since v0.7.1 (git diff v0.7.1 v0.7.2 -- db/): 0104_app_restarts (a column
    apps.restart_generation, guard app_restarts_guard) and 0105_agent_model_prices (rates only, no table
    changes, guard agent_model_prices_guard). The operator runs lake migrate from the image before the new
    binary serves (orama-operator#936), so nothing is vendored. Without 0105 a Lake with billing on refuses the
    gpt-6.x models and gpt-5.5.
  • Settings: nothing new is read from the environment (git grep getenv between v0.7.1 and v0.7.2 adds
    nothing); environment_variables() and stripe_webhook_events are unchanged. The new refusals are Lake's own
    and need no setting: SQL writes and over-cap reads on the App gateway, a service_tier other than default,
    a model whose rates are not published while billing is on, and a workspace model that is no longer served.
  • Agents follow the workspace's choice through the gateway; no agent image changes. The sets name
    agents.model: gpt-6.1-sol, which only says what agents ask for.
  • Restart replaces pods only with orama-operator#1033,
    an operator patch released separately. Either may roll first: the operator reads the column tolerantly.

Known issues

  • Security: a confirmed or auto-fix repair or change turn is checked against its app when it starts, but its
    task credential's write scopes are capability-level, so it could act on another app or repository of the same
    workspace the person may write (#1216). Auto-fix now also
    starts on app_errors, so this applies to more turns.
  • Security: Lake's read-only SQL guard and Zetta's statement parser disagree on comments and statements
    (#926); the App gateway's sql goes through that guard.
  • The other open security issues listed in v0.7.0 remain open.
  • app_logs is empty on nonprod: no Loki-compatible log source is configured there
    (#1206), so the app_errors detector reads nothing there.
  • The gateway was not exercised against OpenAI by the pull requests' suites (they use a fake provider); the dev
    environment's gate checked the workspace model setting, not a paid long prompt.
  • The end-to-end failures from v0.7.1's gate on testing are not fixed here: Ask Orama ranks the wrong at-risk table
    (#1225), walkthrough.mjs fails after the Quickstart's failure
    path (#1226), and the Experiments Agent's turn crashes
    (agent-experiments#26). Also open: a file dropped
    before a new workspace is ready is never loaded (#1224), and
    the v0.7.0 and v0.7.1 follow-ups listed on their pages.

Where it has run

lake-dev ran dev-11153627, the commit this release is cut from, and passed the dev gate: health, an App's
gateway SQL with its read binding, the 5,000-row cap, and the workspace model setting. The release image goes to
testing first, where the e2e gate runs.

Image: oramasearch/lake:v0.7.2, also main-ef4a519f, digest
sha256:b05e5fe553fe16e520903f2496d5e063fb9c8c94623092fb8edd52f2b8df6e63, built on the cluster BuildKit (linux/amd64)
from a detached checkout of the tag (ef4a519f, LAKE_BUILD_REVISION) against zetta v0.5.0 and liquid-bridge's
liquid pin (27d56ba). The Docker Hub repository stays private.

Verification

On the release branch (dev at 11153627 plus the version and these notes, defd7172), macOS arm64, Release,
against zetta v0.5.0 (e11c02f5) and liquid at liquid-bridge's pin (27d56ba), PostgreSQL (a database of its own),
LAKE_TEST_SKIP_STALE_SWEEP=1, a live Zetta v0.5.0 node (LAKE_TEST_ZETTA_URL, LAKE_TEST_ZETTA_TOKEN,
LAKE_TEST_ZETTA_BINARY), ctest -j6 (248 s):

Check Result
Build every target builds and links
Full suite 1,658 tests: 1,642 passed, 0 failed, 16 skipped, 1 disabled
Live Zetta 12 tests ran and passed against the real engine: QuickstartZetta.* (3), WorkspaceStoreZetta.* (6), WorkspaceRecordsZetta.*, RequestRecorderZetta.* and FeaturePlanZetta.*; none skipped
Skipped the PostgreSQL variants of six `HttpResponseIsol...
Read more

Release v0.7.1

Choose a tag to compare

@homer6 homer6 released this 26 Oct 03:42
7241d54

Released: 2026-10-25 · Branch: main · Previous: v0.7.0 (2026-10-24) ·
Tracking: #1219, epic #1158.

A patch:

  • The Quickstart decides what to build with you. Its model step is an Ask Orama conversation on the
    dataset you dropped. It trains on any column types, a clustering model trains, and every state after the
    conversation says what happens next. The website code comes as a copy-ready agent prompt.
  • A failed app tells you, and the Application Agent fixes it. A failed rollout or build is a
    notification with Fix with the Application Agent; an app can be changed with instructions, and auto-fix
    is opt-in. The Apps page counts Degraded apps as needing attention.
  • Ask Orama's confirm card is readable, and it never proposes a change that cannot work.
  • Auto-Suggest finds opportunities again, and a scan survives leaving the page.

No migrations and no new environment variables since v0.7.0. Ships with agent-ask-orama v0.2.1 and
agent-application v0.2.1 (Before it rolls).

Features

PR Change
#1199 The Quickstart's model step is an Ask Orama conversation (#1182). Once search works, Decide What to Build opens Ask Orama on the dataset: its name, row count, each column's kind and whether it can be predicted, and the index. It leads with ideas: a model, a target derived from a column (a threshold, bands or a set of values), M-3 clustering, Auto-Suggest opportunities, an analysis, or an app built by the Application Agent (such as a dashboard). Confirming a model trains it, and the steps continue. Nothing to Predict Yet is gone; Choose the columns myself stays as a fallback. New capability start_quickstart_conversation. quickstart_status names the conversation. train_quickstart gains derived_target and methodology: "M-3" (a clustering model is not published for the website: zetta serves predictions only for gradient-boosted models). The conversation lives in the workspace Zetta (../api/quickstart.md)
#1186 The Quickstart trains on any column types, detected automatically, and never fails on transform_required (#1181). The training copy reads up to 1,000,000 rows and must fit upload_dataset's 16 MiB (#1190)
#1187 Website code as a Copy agent prompt tab (the default), carrying the source and everything an agent needs to install it, beside a Source code tab (#1185)
#1218 A failed app is a notification, fixed or changed with the Application Agent, with opt-in auto-fix (#1207). A background pass records a notification in the app's workspace Zetta when a rollout ends Degraded, Failed or crash-looping at the desired generation, or the newest build fails; Ready again resolves it. The Apps list shows a banner per failed app; the app page has Fix or Change This App (open notifications, Fix with the Application Agent, a Change this app box, and Auto-fix: off by default, 1 to 10 attempts, default 3). The agent proposes a diff; once confirmed it commits through commit_repository_files (liquid drips, attributed to the person) and redeploys. Auto-fix makes one attempt per failed generation as the person who turned it on, then stops and says so. Account → Notifications lists every notification of the workspace. New capabilities list_notifications, dismiss_notification, get_auto_fix, set_auto_fix, repair_app, change_app; agent_turn takes kind: repair or change for agent-application. Records: lake_resource_notifications and lake_resource_repairs in the workspace Zetta; nothing in Postgres (../api/notifications.md)

Fixes

PR Change
#1213 A Quickstart clustering model trains, and every state after the conversation has one next step (#1203, #1204). Lake no longer sends a held-out share for M-3, which zetta refuses; the engine chooses the number of segments (2 to 8). After a confirm the page switches to the training at once and says how far along it is; a refused or failed training gives the reason in plain words with Try again (same choices) and Ask Orama to fix it. A confirmed train_quickstart writes what started into the conversation, or ok: false, code: training_failed (new error code) with the plain reason, instead of "Done: train_quickstart.". The conversation page carries quickstart_state, model, model_problem and quickstart_next so Ask Orama can follow it (../api/quickstart.md, ../ui/quickstart.md)
#1211 The Apps page counts Degraded apps as needing attention (#1208). Failed becomes Needs attention (Failed or Degraded), linked to ?filter=attention, with Show all apps back (../ui/apps.md)
#1189 Auto-Suggest suggests opportunities again: plan_features works from a dataset alone, with its snapshot cutoff. With agent-autosuggest v0.2.0 (#24), Scan my workspace keeps its Opportunities (#1153; e2e T6 passes on dev)
#1201 A scan survives leaving the page (#1195). A finished scan's Opportunities are saved when the agent publishes them, not when someone reads them. list_opportunities returns last_scan, so the page shows Scanning… started 2m ago or the results without ?task=. A second scan while one runs joins it (../api/discover.md)
#1198 Ask Orama's proposals:
• the confirm card shows each argument as name, value and description, wrapping across the full width, with long values folded (#1192);
• a proposal naming a project that does not exist is withheld before it is shown, and a new project's pipeline is one confirmable compile_pipeline with create_project: true (#1193);
• every turn carries Lake's record of the conversation's proposals (pending, confirmed, failed, withheld), so after a failure "create it then" gets a new, corrected proposal (#1194) (../api/ask-orama.md)
#1217 A test pins Lake's half of the App Zetta binding contract (#1205). HttpResponseIsolation.AnAppsZettaBindingReadsItsWorkspaceThroughLake builds the binding as the operator renders it after orama-operator#1001 and proves the app's read of its own workspace is served, and another workspace, an action and a wrong token are refused. No Lake code changes (../api/workspaces.md)

Before it rolls

  • Schema: unchanged since v0.7.0 (git diff v0.7.0 v0.7.1 -- db/ is empty). The operator runs lake migrate
    from the image, which applies nothing new.
  • Settings: nothing new is read from the environment (git grep getenv between v0.7.0 and v0.7.1 adds only
    the test variables LAKE_TEST_ZETTA_URL and LAKE_TEST_ZETTA_TOKEN); environment_variables() and
    stripe_webhook_events are unchanged. The new refusals are Lake's own (a repair turn naming another app, auto-fix
    past its attempts, a clustering of fewer than 2 rows) and need no setting.
  • Agents: roll with agent-ask-orama v0.2.1 (prompt v17: it follows the Quickstart's state after a confirm,
    agent-ask-orama#56) and agent-application v0.2.1
    (repair and change turns, agent-application#40; without
    it a repair turn gets the old new-app behaviour). The other agents stay at v0.2.0.
  • Apps with workspace data start only once orama-operator#1001
    (the binding as JSON) is released and rolled (#1205).

Known issues

  • Security: a confirmed or auto-fix repair or change turn is checked against its app when it starts, but its
    task credential's write scopes (commit_repository_files, build_app, set_app_env, deploy_app) are
    capability-level, so it could act on another app or repository of the same workspace the person may write
    (#1216).
  • Those of [v0.7.0](https://github.com/oramasearch/lake...
Read more

Release v0.7.0

Choose a tag to compare

@homer6 homer6 released this 25 Oct 05:15
7fa1e51

Released: 2026-10-24 · Branch: main · Previous: v0.6.1 (2026-10-24).

The first release after lake.live's launch, gathered in #1158.
A new home page, Quickstart, takes a dropped dataset to search and model answers on your own
website. A browser key is the public credential that website holds. Orama can give credits
and run coupon codes for ads. Signing out no longer revokes API keys. A white-labelled Lake
gets its customer's domain, sign-in and brand. lake migrate and lake manifest let
the image apply its own schema. Plus security fixes and three memory bugs found by AddressSanitizer.

Four migrations (0100 to 0103), all additive. The schema reaches the control database before
the image rolls (Before it rolls).

Features

PR Change
#1157 Quickstart, the new home page at / (Ask Orama moves to /ask; old links redirect). Drop a CSV or JSONL file: it becomes a dataset, a published search index, website code (plain page, React, curl) with a live search box, then Train a Model trains and publishes a model whose answers show beside each result. start_quickstart, quickstart_status, train_quickstart and quickstart_snippets on REST, MCP and the agent documents (Start Here). Each step calls the existing capabilities as the person, so permissions, the balance gate, audit and metering are theirs; progress is kept in the workspace Zetta (../api/quickstart.md, #1128)
#1177 Browser keys (lbk.<id>.<secret>): the credential a website's page holds. Bound to one workspace, one index and optionally one model; only search_index and predict on those; only from its listed origins (CORS answers those alone); a per-minute limit shared by every replica (600 by default). Not tied to any session: it lasts until revoked or its expiry. create_browser_key, list_browser_keys, revoke_browser_key; the Quickstart makes one in one click; Search Index gets Website Keys. Migration 0102_browser_keys, hashes only (../api/browser-keys.md, #1152)
#1155 Credits from Orama. An instance administrator (or the Lake's admin token) gives an account a credit and revokes what is left of it: grant_balance_credit, revoke_balance_credit, console Give credit. Spent before paid money, earliest expiry first; never unlocks the overdraft; when it expires or is revoked only its unused part stops counting. One email, audited centrally. Migration 0100_balance_grants (../api/billing.md#credits-from-orama, #1134)
#1165 Coupon codes for ads. create_coupon, list_coupons (redemptions, first top-ups and spend per campaign), disable_coupon, redeem_coupon; Administration → Coupons. A redemption is an ordinary credit from Orama. lake.live/?code=… carries the code through the region picker and sign-up. One redemption per account and per email (kept as a hash), new accounts only by default, failures rate-limited. Each region creates its own codes (#1159). Migration 0101_coupons (../api/billing.md#coupon-codes, #1156)
#1147 lake migrate applies the schema built into the image (one transaction, an advisory lock, JSON report, exit codes) and lake manifest prints what the image needs from the operator: the schema, its Stripe webhook events and products, and every environment variable it reads. server_info lists the Stripe events. With orama-operator#965 a lake release is only an image change (../deployment/operator-interface.md, #1119)
#1167 Sign-in on a white-labelled Lake's own domain: LAKE_HOST_ALIASES (browser pages on the Orama host redirect to the customer's), LAKE_CONTROL_HOST, and LAKE_SIGN_IN_SSO_PROVIDER, which makes the customer's identity provider the only way in, refusing Lake's other sign-in methods in its own pool. All off until set (../deployment/white-label-sign-in.md, #1136)
#1174 A white-labelled Lake shows its customer's brand (LAKE_BRAND_* or LAKE_BRAND_FILE): the console, sign-in, lake's own pages, every email, llms.txt, the skill (/skills/<short_name>/SKILL.md), MCP initialize, tools/list, OpenAPI and server_info. Unset, lake.live is byte-for-byte unchanged. The brand's terms and privacy links win over LAKE_TERMS_URL / LAKE_PRIVACY_URL (../serving/brand.md, #1135, #1137)
#1171 Terms and privacy links under every sign-up and sign-in form, Clerk's and Lake's own (LAKE_TERMS_URL, LAKE_PRIVACY_URL; default Orama's published pages; off for none; a malformed URL stops lake). viewer.legal (liquid-bridge#15)
#1139 The lake.live plan cards list "Your own domain, white-labelled (available on request)" on Private Warehouse, Dedicated Cluster and On-Prem (#1129)
#1130 The apex lake.live answers llms.txt, llms-full.txt and the skill with every region's MCP server and guides, instead of the picker's HTML (#1110)

Security and identity

PR Change
#1173 Signing out ends only that browser. A Clerk sign-out used to revoke every session, API key and device of the account. Lake now binds each browser session to its Clerk session (the default session token's signed sid), and a session event ends only the browser it names. No sign-out ends a key, including keys made in that browser. A new Sign out everywhere (sign_out_everywhere, Sessions page) ends everything for your own account. Bans, locks, deletion and revoke_user_sessions still end everything. Migration 0103_clerk_session_binding (#1163, #1172)
#1138 Under Clerk, Lake's own password sign-up and sign-in in the lake pool are refused (403), by the same rule that makes viewer say password_sign_up: false; it was possible to skip Clerk's sign-up control (#1098, the provisioning step of LSR-011)
#1133 An unknown email takes as long to reject as a wrong password (a dummy Argon2id check), so response times no longer show which emails have accounts (#1101)
#1131 Per-client limits believe X-Real-IP / X-Forwarded-For only from trusted proxies (LAKE_TRUSTED_PROXY_CIDRS; unset: loopback and the private ranges, so nothing changes before the operator sets it) (#1083)
#1169 A proxy listener (LAKE_PROXY_PORT): client-address headers are believed only on the port the ingress alone can reach, and the main port then trusts no one by default. Off until the operator sets it (orama-operator#961) (#1132)

Fixes

PR Change
#1178 Three memory bugs from the AddressSanitizer run: a segfault at start with no control store (#1168), a heap-use-after-free in the memory store's SCIM group upsert (#1175), and a container read after unlock in the search job thread (#1176)
#1162 The tools/e2e/ browser scripts drive the React console (Clerk ticket, first top-up on Stripe's test Checkout, the workspace), so the dev → testing gate's browser half runs again (#1099)
#1160 Docs: lake.live's complete price list as Steve approved it (23 rates, effective 2026-11). Not published by this release: set_price_list on each region is a separate step

Before it rolls

  • The schema through 0103 reaches each control database before the image. Either
    orama-operator#965 (the operator runs lake migrate from this image, the first that has it), or an
    operator release that vendors db/schema.sql through 0103 with clerk_session_binding_guard
    selected. A v0.7.0 pod on the v0.6 schema fails Clerk sign-in, grants, coupons and browser keys.
  • No new required setting. Every new variable (LAKE_PROXY_PORT, LAKE_TRUSTED_PROXY_CIDRS,
    LAKE_TERMS_URL, LAKE_PRIVACY_URL, LAKE_HOST_ALIASES, LAKE_CONTROL_HOST,
    LAKE_SIGN_IN_SSO_PROVIDER, LAKE_BRAND_*) defaults to today's behaviour; lake manifest
    declares them all.
  • After the image: orama-operator#961 (the proxy port) and #963 (a customer's domain and brand)
    roll only once every Lake runs v0.7.0.
  • Behaviour to know: revoking a device no longer ends keys made in a browser on it; use
    Sign out everywhere to end every credential.

Known issues

  • Security, open: 76 issues labelled security remain open in lake, from the pre-launch review
    and its companions. LSR-200 (#1006, hosted code
    spoofing forwarding headers) is closed once orama-operator#961 routes the i...
Read more

Release v0.6.1

Choose a tag to compare

@homer6 homer6 released this 24 Oct 07:27
cd5dca1

v0.6.1

Released: 2026-10-24 · Branch: main · Previous: v0.6.0 (2026-10-24).

A patch so a blocked account can stop or rename its workspace (#1122).
Everything else is v0.6.0. No schema change: it rolls by image alone.

Fixes

PR Change
#1125 The balance gate refused update_workspace as a whole, so an account blocked by its balance (every account before its first top-up, including those created before v0.6.0 with a running workspace) could delete its workspace but not stop it. It is now refused only with desired: running; stopping or renaming always passes (../api/billing.md)

Before it rolls

Nothing new: operator v0.9.5 or later, as for v0.6.0.

Known issues

Those of v0.6.0, and auto top-up needs each Stripe key's PaymentIntents
permission (#1123, #1124).

Verification

On the release branch (dev at f19d603f plus the version), macOS arm64, Release, static, against zetta
v0.5.0 (e11c02f5) and liquid 27d56ba, PostgreSQL 17, ctest -j8 (269 s): 1,481 tests, 1,461 passed,
0 failed, 19 skipped
(the same as v0.6.0: a live Zetta, the Hugging Face client, a stdio MCP runner, a
browser), 1 disabled (SearchBenchmark). New: Balance.ABlockedAccountCanStopOrRenameAWorkspaceButNotStartOne.

Release v0.6.0

Choose a tag to compare

@homer6 homer6 released this 24 Oct 06:38
20c8016

v0.6.0

Released: 2026-10-24 · Branch: main · Previous: v0.5.2 (2026-10-24).

lake.live becomes prepaid: each account has a balance, tops it up through Stripe, is warned and
emailed as it runs low, and cannot start paid work once it is $50 below zero
(#1114, liquid-bridge decision 67). Also: people
whose Clerk profile has no name can sign in, and the sign-in page no longer loops on an error.

Read Before it rolls. Migration 0099 and new Stripe events need an
operator release first.

Features

PR Change
#1118 Prepaid account balance (billing on only). Balance = top-ups − charges, in Postgres with billing (decision 56). Top-ups through Stripe Checkout: the first exactly $20, later ones $20 or more; each payment credited once from the webhook and audited. Charges refreshed every 5 minutes from each workspace's usage at the price list; a failed usage read holds the balance. 51 capabilities that start paid work answer 402 payment_required (balance, limit, Add funds link) when the balance is more than $50 below zero, or before an account's first top-up; reads, sign-in, stops and deletes stay open (PaidWorkContract fails on an unmarked one). A banner on every page, a warnings entry on REST and MCP answers, and emails (near the limit, below $0, blocked) sent once per crossing, all linking to /billing/add-funds. Auto top-up, off by default: a card saved through Stripe, charged off-session once per crossing, at most 5 a day; a decline turns it off and emails the owner. An issued invoice for a topped-up account is paid from the balance. Migration 0099_account_balances. Settings: LAKE_BALANCE_FIRST_TOP_UP_USD (20), LAKE_BALANCE_MINIMUM_TOP_UP_USD (20), LAKE_BALANCE_OVERDRAFT_USD (50), LAKE_BALANCE_WARNING_USD (10), LAKE_BALANCE_REFRESH_SECONDS (300), LAKE_AUTO_TOP_UPS_PER_DAY (5) (../api/billing.md)

Fixes

PR Change
#1117 Sign-in with no name in Clerk (#1115): the lake template's name is null for a Clerk user with no first or last name, and lake refused the token. A null name or email now counts as missing. A refused token logs clerk_sign_in_refused with the reason (never the token). After an error the sign-in page stops retrying and offers Try again. The invitation wording appears only where the Clerk instance's sign-up is restricted (#1116)
#1107 Who owns the Microsoft sign-in app and what blocks its publisher verification (docs, #756)

Before it rolls

  • An operator release that vendors db/schema.sql through 0099_account_balances (guard
    account_balances_guard) and adds the Stripe events the balance needs to lakeStripeWebhookEvents:
    setup_intent.succeeded, payment_intent.succeeded, payment_intent.payment_failed, with
    checkout.session.completed and checkout.session.async_payment_succeeded on every billing Lake.
  • The Lake's restricted Stripe key (appKey) must be allowed to write PaymentIntents (auto top-up).
  • Existing accounts start at $0: on a billing Lake they must top up before starting paid work.
  • Each Lake restarts once.

Known issues

  • No live Stripe top-up or auto top-up has run yet; the first is the dev test after the roll.
  • The balance warning adds a few Postgres reads to every API call while billing is on; not measured.
  • Those of v0.5.2: #1083, #1101, #1110, the incomplete published price list.

Verification

On the release branch (dev at 58356286 plus the version), macOS arm64, Release, static, against zetta
v0.5.0 (e11c02f5) and liquid 27d56ba, PostgreSQL 17, LAKE_TEST_SKIP_STALE_SWEEP=1, ctest -j8
(383 s):

Check Result
Build every target builds and links
Full suite 1,480 tests: 1,460 passed, 0 failed, 19 skipped, 1 disabled
Skipped they need a live Zetta (WorkspaceStoreZetta ×6, WorkspaceRecordsZetta, RequestRecorderZetta, and the agent and Agent Factory HttpResponseIsolation cases ×8), the Hugging Face Python client, a stdio MCP runner, or a browser (InstanceOrigin); SearchBenchmark is disabled by design
Web app npm run build (391 tests) and ./tools/ui/test-token-gates.sh pass (#1118); 387 tests with #1117

Not yet run: a live Stripe top-up and auto top-up (tested on dev after the roll), the Linux BuildKit
lane (the image build is the Linux build), and the testing e2e gate (lake-e2e has no image).

Release v0.5.2

Choose a tag to compare

@homer6 homer6 released this 24 Oct 04:32
69d2519

v0.5.2

Released: 2026-10-24 · Branch: main · Previous: v0.5.1 (2026-10-24).

A patch so people can sign in to lake.live with Google: lake refused any request whose headers
passed about 4 KiB with 431 header_too_large, which a browser signed in through Clerk reaches after
an OAuth round trip (#1108). Everything else is v0.5.1.

Changes

PR Change
#1109 Request headers up to 1 MiB in all (LAKE_HTTP_MAX_HEADER_BYTES, 8 KiB to 16 MiB). boost::http's defaults were 8 KiB in all and 4 KiB a field, and its parser applies the field limit to the whole header block, so every request's headers were capped near 4 KiB; lake now sets both to the limit. An idle connection costs about 49 KiB of resident memory (was about 25 KiB) (../serving/http.md)
#1106 How a person gets an account on lake.live (docs)
#1105 v0.5.1's notes name its image digest (docs)

Before it rolls

  • Nothing new: no schema change, no new required setting. orama-operator v0.9.2 or later, as for
    v0.5.1.
  • At lake's default 4096 connections, idle connections can take about 200 MiB; lake.live's pods are
    limited to 512 MiB.

Known issues

Those of v0.5.1, and:

  • The apex lake.live answers /llms.txt with the region picker, not the agent text
    (#1110).
  • lake.live's published price list has only the model-token rates (lake#755).

Verification

  • HttpConnections.ABrowsersClerkCookiesFitLakesHeaderLimits: a 12 KiB Cookie with 20 KiB of headers
    is answered; past 1 MiB, one field or many, it is still 431.
  • HttpConnections.AnIdleConnectionCostsLittleMemory, now with lake's own limits: 49 KiB per idle
    connection, under its 256 KiB bound.
  • All 216 HTTP tests pass (ctest -R Http, macOS, against zetta v0.5.0). The full suite was not
    rerun for this patch
    (launch day; Steve: "cut the new version of lake").