v3.1.0

Research only the files you choose, with every claim checked

• Choose files to research. Tick files and folders on the Corpus page and ask; the research uses only those files, at the versions they had when it started, and the run page and the verification bundle list them.
• Every claim is checked before you see the report. In research limited to your corpus, each claim is checked against the passages its citations point to. Unsupported passages are rewritten or removed, and the run page shows how accurate the checking model was measured to be.
• Folders and many more file kinds: Word, OpenDocument, RTF, spreadsheets, PowerPoint, EPUB, notebooks, LaTeX, BibTeX and source code. Citations open at the cited page, sheet, slide, cell or line.
• Offline: research refuses to start unless every model runs at an address on this computer.
• A corpus per project on the desktop. Documents from earlier releases move to "Documents from before 3.1", and the original is kept.

The desktop app is published for Apple Silicon Macs only. Windows and Linux users keep v3.0.3, which has none of the above.

macOS first launch: the app is not signed with an Apple Developer ID. Drag it to Applications, then run xattr -dr com.apple.quarantine "/Applications/Research Assistant.app" in Terminal once.

Checking every claim makes research on your files slower and costlier. It was measured against a model judge, not a person, and on two routings. Read the known limitations before relying on it.

---

Multi-arch container images (linux/amd64, linux/arm64) pushed to GHCR:

• ghcr.io/adityamhaske/multi-agent-research-assistant-api:3.1.0
• ghcr.io/adityamhaske/multi-agent-research-assistant-worker:3.1.0
• ghcr.io/adityamhaske/multi-agent-research-assistant-frontend:3.1.0

latest points at this release.

Deploy the full stack on a free Oracle Cloud Ampere VM — see deploy/README.md.

What's Changed

• docs(internal): V3.1 Corpus RFC and audit by @adityamhaske in #212
• fix(corpus): only a corpus-only run can read the corpus by @adityamhaske in #213
• feat(corpus): enforce a run's selection at the corpus port by @adityamhaske in #214
• feat(runs): freeze a corpus run's file selection and enforce it on both hosts by @adityamhaske in #215
• feat(runs): check, expose and bundle a selected-files run's provenance by @adityamhaske in #217
• feat(runs): the Offline guarantee, enforced in the engine on both hosts by @adityamhaske in #219
• fix(desktop): each project has its own corpus; the shared store moves to a project of its own by @adityamhaske in #220
• feat(grounding): the claim-level validator for selected-files runs by @adityamhaske in #221
• feat(grounding): repair, gate and record for selected-files runs by @adityamhaske in #222
• feat(corpus): one format registry, canonical extraction record, archive limits by @adityamhaske in #223
• feat(corpus): Tier A formats: source code, LaTeX/BibTeX, CSV/TSV, notebooks by @adityamhaske in #224
• feat(corpus): Tier B formats: Word documents, Excel and OpenDocument spreadsheets by @adityamhaske in #225
• feat(corpus): Tier C formats: PowerPoint, OpenDocument text, EPUB, RTF by @adityamhaske in #226
• feat(corpus): folders: stable ids, membership, folder uploads, folder selections by @adityamhaske in #227
• feat(corpus-ui): folder tree, cross-folder selection, folder drop, folder uploads by @adityamhaske in #228
• feat(research-ui): ask about selected files, the Offline switch, and what a run used by @adityamhaske in #229
• feat(citations): open a corpus citation where it is cited by @adityamhaske in #230
• docs(user-guide): research over the files you choose by @adityamhaske in #231
• release-audit(v3.0.2): the 48-hour watch closed with no reports (M22.05) by @adityamhaske in #207
• chore(deps): bump @vitest/mocker and vitest in /frontend by @dependabot[bot] in #209
• docs(user-guide): mark the corpus guide as not in a released version yet by @adityamhaske in #232
• release(v3.0.3): offer the published installers by @adityamhaske in #233
• ci(desktop): drive the engine inside the macOS disk image through both gates by @adityamhaske in #234
• release-audit(v3.0.3): post-release — the live site offers 3.0.3 by @adityamhaske in #235
• release-audit(v3.0.3): the macOS packaged checks, after the release Mac restarted by @adityamhaske in #236
• fix(desktop): run limits default to the server's, and Settings sees the saved endpoint by @adityamhaske in #237
• test(corpus): project A's documents are unreachable from project B, on both hosts by @adityamhaske in #239
• release-audit(v3.0.3): the 48-hour watch closed with no reports (M22.05) by @adityamhaske in #238
• test(selection): an injected read, the search cache, and one- and two-file selections by @adityamhaske in #241
• fix(grounding): a run limit that stops checking says which one, and by how much by @adityamhaske in #240
• feat(citations): open a cited PDF at its page; desktop previews and downloads authenticate by @adityamhaske in #242
• feat(runs): reopen a cited document wherever it lives now, and nothing else by @adityamhaske in #243
• feat(ci): every frozen engine reads every shipped format as the source tree does by @adityamhaske in #245
• fix(desktop): --fake embeds in-process, as the server does under fake mode by @adityamhaske in #244
• feat(citations): a citation opens through the run or session that made it by @adityamhaske in #246
• fix(desktop): the legacy migration's record names the corpus it moved by @adityamhaske in #247
• eval(retrieval): selected-files leakage is 0, and the per-file cap doubles passage recall by @adityamhaske in #249
• test(upgrade): a real v3.0.2 desktop moves, downgrades and survives a failed move by @adityamhaske in #248
• feat(gates): G-4, a 3.1 bundle under the unchanged v3.0.2 verifier by @adityamhaske in #250
• feat(gates): G-3, the released 3.0.2 app opens 3.1 data by @adityamhaske in #251
• fix(desktop): the pre-3.1 corpus move no longer crashes the Windows app at launch by @adityamhaske in #253
• eval(grounding): the G-6 claim set, 425 claims over every shipped format by @adityamhaske in #252
• eval(checking-cost): the G-5 runner, and the check's wall time in its report by @adityamhaske in #254
• eval(gates): G-6 reports per kind, source count and numbers; G-5 counts retries by @adityamhaske in #255
• release-audit(v3.1.0): G-3 evidence from CI on all three OSes by @adityamhaske in #256
• eval(checking-cost): G-5 measured — §9 multiplies p95 time by 6.4x by @adityamhaske in #258
• eval(grounding): G-6 reference labels by Claude Sonnet 4.5 as a judge (amendment A-1) by @adityamhaske in #257
• test(grounding): AC-42 — no path around the gate by @adityamhaske in...

v3.0.3

V3.0.3

The desktop app researches on a local model with no API key, as Settings always said it could.

• Desktop research on Ollama runs with no API key saved. Every desktop release since 2.1.0 refused it at once with "No provider key found", although the server and the command line never needed a key for Ollama (#202).
• Settings → Models on the desktop offers a local model whenever Ollama is running, and saving a Custom Endpoint base URL saves it instead of answering "Unknown provider." The run form's Local LLM now defaults to a model large enough to return citable evidence, not one marked chat only (#202).
• Successful sign-ins no longer count toward the sign-in limit. Five failed sign-ins still lock the account for 15 minutes, the right password included (#200).
• Settings → Appearance loads without a hydration error, and the public site no longer requests /api/v1/projects on every page (#201, #205).
• Corpus-only mode is described as what it enforces: evidence only from your uploaded documents, web search off, indexing on a local model. The question and the excerpts still go to whichever provider each research model uses; a local indexing model does not make a run offline (#210).

macOS: the app is ad-hoc signed, not notarized. On first launch macOS says it could not verify the app; open it once with:

xattr -dr com.apple.quarantine "/Applications/Research Assistant.app"

Known issues — the full list is under Known in the changelog. The ones most likely to matter:

• Corpus-only mode does not stop model calls from leaving your machine unless you have chosen local research models.
• On the desktop app every project shares one corpus, so corpus-only research and chat's "My corpus" scope draw on documents uploaded under any project.
• This release's bundle verifier needs Python 3.14; on an older Python, 3.0.1's verifier checks the same bundles.

---

Multi-arch container images (linux/amd64, linux/arm64) pushed to GHCR:

• ghcr.io/adityamhaske/multi-agent-research-assistant-api:3.0.3
• ghcr.io/adityamhaske/multi-agent-research-assistant-worker:3.0.3
• ghcr.io/adityamhaske/multi-agent-research-assistant-frontend:3.0.3

latest points at this release.

Deploy the full stack on a free Oracle Cloud Ampere VM — see deploy/README.md.

What's Changed

• fix(auth): the per-account login limit counts failures, not successful logins by @adityamhaske in #200
• fix(ui): the theme picker no longer breaks hydration on every load of /settings/appearance by @adityamhaske in #201
• release(v3.0.2): offer the published installers by @adityamhaske in #203
• release-audit(v3.0.2): post-release — the live site offers 3.0.2 by @adityamhaske in #204
• fix(desktop): research on Ollama with no API key, as Settings says it can by @adityamhaske in #202
• fix(site): public pages stop fetching /api/v1/projects by @adityamhaske in #205
• release(v3.0.2): the desktop keyless fix and the site fetch fix are on main by @adityamhaske in #206
• chore(deps): bump vite from 8.1.5 to 8.3.5 in /frontend by @dependabot[bot] in #208
• release(v3.0.3): prepare the patch — the four fixes 3.0.2 promised by @adityamhaske in #210
• release(v3.0.3): record the eval waiver by @adityamhaske in #216
• release(v3.0.3): account for the record-only PRs and pre-tag warnings by @adityamhaske in #218

Full Changelog: v3.0.2...v3.0.3

Desktop bundles for v3.0.3

These builds are not code-signed. Your OS will warn you on first launch:

OS	What you will see	What to do
macOS	A block on first launch — the app is not signed with an Apple Developer ID or notarized	Drag the app to Applications, then run xattr -dr com.apple.quarantine "/Applications/Research Assistant.app" in Terminal once
Windows	SmartScreen warning	More info → Run anyway
Linux	AppImage is not executable	chmod +x *.AppImage (.deb needs nothing)

Verify your download against SHA256SUMS before running it:

sha256sum -c SHA256SUMS --ignore-missing

Research bundles from v3.0.0 onward are format v2, and a verifier from before
v3.0.0 refuses them. Check a bundle with the verify_bundle shipped in this
release; it still verifies every earlier (v1) bundle, unchanged.

v3.0.2

V3.0.2

Research through your own model router runs for real, and a reworked report's counts describe the report you approved.

• Custom, OpenRouter and Ollama setups on a self-hosted server research your question. ./start.sh recognised only Google, Anthropic and OpenAI keys, so any other setup started in demo mode and every question came back as the same scripted report (#171, #177).
• Approving a plan no longer fails the run on OpenAI-compatible and Anthropic models (#171, #177).
• A rework no longer doubles a run's evidence, and "cited" counts against the revision you approve (#181).
• The executor searches instead of scraping search-engine results pages (#178). Measured before shipping: citation support 96.9% over 190 judged claims on ten fixed questions (Gemini 2.5 Flash, judged by Claude Sonnet 4.5); one run, so not evidence of a change from 3.0.0's 96.4%.
• Every build says which commit built it. The published server images report this tag's commit, and an image you build reports your checkout's (#170, #188).
• Python 3.14 for the server and the desktop engine, Node 26 for the web image and CI, and Next.js 16.4.2 with updated sharp and source-map-js, past high-severity advisories.

macOS: the app is ad-hoc signed, not notarized. On first launch macOS says it could not verify the app; open it once with:

xattr -dr com.apple.quarantine "/Applications/Research Assistant.app"

Known issues — the full list is under Known in the changelog. The ones most likely to matter:

• Desktop with Ollama and no API key: research fails at once with "No provider key found", and Settings offers no local model. Until the next release ships the fix (#202, merged), save any text as the Custom Endpoint key in Settings → Connections — Ollama ignores it.
• More than five successful sign-ins to one account within 15 minutes lock it for the rest of the window (fixed on main, #200).
• This release's bundle verifier needs Python 3.14; on an older Python, 3.0.1's verifier checks the same bundles.

---

Multi-arch container images (linux/amd64, linux/arm64) pushed to GHCR:

• ghcr.io/adityamhaske/multi-agent-research-assistant-api:3.0.2
• ghcr.io/adityamhaske/multi-agent-research-assistant-worker:3.0.2
• ghcr.io/adityamhaske/multi-agent-research-assistant-frontend:3.0.2

latest points at this release.

Deploy the full stack on a free Oracle Cloud Ampere VM — see deploy/README.md.

What's Changed

• release(v3.0.1): offer the published installers by @adityamhaske in #169
• A mandatory release process, and an audit that runs it by @adityamhaske in #170
• Fix gated runs failing on OpenAI-compatible routes, and every question getting the demo report by @adityamhaske in #171
• Reset an agent's instructions to the shipped prompt from inside its editor by @adityamhaske in #179
• An approval is not a 500 when its report cannot be filed into project memory by @adityamhaske in #180
• fix(runs): a rework stores no second copy of the evidence, and the run page counts the latest draft by @adityamhaske in #181
• Prove real-mode worker runs end to end, make loop ownership explicit, and label demo runs plainly by @adityamhaske in #177
• Refuse search-engine results pages in read_webpage and steer the executor to web_search by @adityamhaske in #178
• chore(deps-dev): bump brace-expansion from 1.1.18 to 1.1.21 in /frontend by @dependabot[bot] in #176
• chore(deps): bump undici from 8.10.0 to 8.11.2 in /frontend by @dependabot[bot] in #175
• chore(deps): bump the frontend-minor group in /frontend with 2 updates by @dependabot[bot] in #174
• chore(deps): bump python from 3.11-slim-bookworm to 3.14-slim-bookworm in /backend in the backend-images group by @dependabot[bot] in #173
• chore(deps): bump next from 16.3.5 to 16.3.6 in /frontend by @dependabot[bot] in #182
• chore(deps): bump next from 16.3.6 to 16.3.8 in /frontend by @dependabot[bot] in #185
• chore(deps): bump sharp from 0.35.4 to 0.35.5 in /frontend by @dependabot[bot] in #184
• chore(deps): bump source-map-js from 1.2.1 to 1.2.2 in /frontend by @dependabot[bot] in #183
• release(v3.0.2): prepare the release — notes, record, and the release evaluation by @adityamhaske in #186
• fix(ci): test the server on the image's Python, and fail the commit that splits them (A10.06) by @adityamhaske in #187
• fix(desktop): freeze the engine on the image's Python, so the suite tests both hosts by @adityamhaske in #189
• fix(build): every build reports its own identity or unknown, never a stale stamp's by @adityamhaske in #188
• release(v3.0.2): include #187-#189 in the record, and plan the checks they need by @adityamhaske in #192
• fix(python): the advertised floor is the shipped Python 3.14, and A10.06 holds it there by @adityamhaske in #193
• chore(deps): update numpy requirement from >=1.26 to >=2.5.4 in /backend by @dependabot[bot] in #190
• chore(deps): bump the frontend-minor group across 1 directory with 3 updates by @dependabot[bot] in #191
• chore(deps): bump node from 25-bookworm-slim to 26-bookworm-slim in /frontend in the frontend-images group across 1 directory by @dependabot[bot] in #172
• fix(desktop): a packaged-app check whose bundle verification did not run fails CI by @adityamhaske in #194
• fix(ci): build and test the frontend on the image's Node, and fail the commit that splits them (A10.07) by @adityamhaske in #196
• release(v3.0.2): scope #192-#194, E1-E3 results, and the verifier's Python 3.14 floor by @adityamhaske in #195
• release-audit(v3.0.2): scope the record PRs themselves (#195 and this one) by @adityamhaske in #197
• fix(release-audit): A3.02 judges the site a release commit serves, not a run it cannot have by @adityamhaske in #198
• fix(deps): patch postcss-selector-parser under the typography plugin (GHSA-rj75-hqrm-r3gf) by @adityamhaske in #199

Full Changelog: v3.0.1...v3.0.2

Desktop bundles for v3.0.2

These builds are not code-signed. Your OS will warn you on first launch:

OS	What you will see	What to do
macOS	A block on first launch — the app is not signed with an Apple Developer ID or notarized	Drag the app to Applications, then run xattr -dr com.apple.quarantine "/Applications/Research Assistant.app" in Terminal once
Windows	SmartScreen warning	More info → Run anyway
Linux	AppImage is not executable	chmod +x *.AppImage (.deb needs nothing)

Verify your download against SHA256SUMS before running it:

sha256sum -c SHA256SUMS --ignore-missing

Research bundles from v3.0.0 onward are format v2, and a verifier from before
v3.0.0 refuses them. Check a bundle with the verify_bundle shipped in this
release; it still verifies every earlier (v1) bundle, unchanged.

V3.0.1: the macOS app opens (#168)

V3.0.1

The macOS app opens. Every Mac download before this one — 1.0.1 through 3.0.0 — was reported as "“Research Assistant.app” is damaged and can't be opened".

• Why: those builds carried only the signature the linker stamps on the executable (flags=adhoc,linker-signed, Sealed Resources=none). Nothing signed the app bundle, so signature verification failed. macOS treats a downloaded app with a failing signature as damaged, and offers no Open Anyway.
• Fix: the whole bundle is now signed before the disk image is built. CI mounts the image it is about to publish and fails if the app inside is linker-signed only, fails codesign --verify --deep --strict, or is not an ordinary Gatekeeper policy case (#168).
• Checked on a Mac against this release's DMG:
  • the checksum matches SHA256SUMS;
  • codesign reports "valid on disk";
  • Gatekeeper gives a quarantined copy a plain policy verdict, not the damage error;
  • the engine inside reports d53d645, the tagged commit.
• Still not Developer ID signed or notarized. Open it once with:

  xattr -dr com.apple.quarantine "/Applications/Research Assistant.app"
  

  The same command opens every earlier macOS build.

Research is unchanged from 3.0.0. Everything else listed as known in 3.0.0 still applies (changelog).

---

Multi-arch container images (linux/amd64, linux/arm64) pushed to GHCR:

• ghcr.io/adityamhaske/multi-agent-research-assistant-api:3.0.1
• ghcr.io/adityamhaske/multi-agent-research-assistant-worker:3.0.1
• ghcr.io/adityamhaske/multi-agent-research-assistant-frontend:3.0.1

latest points at this release.

Deploy the full stack on a free Oracle Cloud Ampere VM — see deploy/README.md.

What's Changed

• release(v3.0.0): offer the published installers by @adityamhaske in #166
• Releases page version index and summaries, homepage features, docs brought up to 3.0.0 by @adityamhaske in #167
• Fix macOS "damaged" error: sign the whole app bundle; cut 3.0.1 by @adityamhaske in #168

Full Changelog: v3.0.0...v3.0.1

Desktop bundles for v3.0.1

These builds are not code-signed. Your OS will warn you on first launch:

OS	What you will see	What to do
macOS	A block on first launch — the app is not signed with an Apple Developer ID or notarized	Drag the app to Applications, then run xattr -dr com.apple.quarantine "/Applications/Research Assistant.app" in Terminal once
Windows	SmartScreen warning	More info → Run anyway
Linux	AppImage is not executable	chmod +x *.AppImage (.deb needs nothing)

Verify your download against SHA256SUMS before running it:

sha256sum -c SHA256SUMS --ignore-missing

Research bundles from v3.0.0 onward are format v2, and a verifier from before
v3.0.0 refuses them. Check a bundle with the verify_bundle shipped in this
release; it still verifies every earlier (v1) bundle, unchanged.

V3.0.0: per-role agent instructions, with what ran recorded (#164)

V3.0.0

Each of the five research agents — planner, executor, critic, synthesizer and follow-up chat — can follow your own instructions instead of the shipped ones, and every artifact records exactly which instructions produced it.

• Settings → Agents edits one agent at a time, starting from the prompt that runs today, up to 2,500 characters. Any agent resets to the shipped prompt on its own.
• The checks that keep research honest cannot be rewritten. Citation verification, contradiction detection, report repair and project chat's refusal line always run on the shipped prompts, whatever an agent is told.
• A run keeps the instructions it started with. An edit made while a run waits at a review gate applies to the next run, not the half-finished one.
• Verification bundles are format v2. They record, for every prompt the pipeline used, its full text, its SHA-256 and whether it was replaced, inside the bundle hash. A verifier from before 3.0.0 refuses v2 bundles. The verifier in this release still verifies every earlier (v1) bundle unchanged.
• Upgrading keeps everything. This release's packaged engine upgraded a data directory written by the real 2.1.0 app:
  • nothing was lost;
  • 2.1.0's bundles still verify;
  • a run waiting at a gate still approves;
  • research recorded before 3.0.0 stays v1.
• Measured by an independent judge. Under the shipped prompts, the fixed ten-question evaluation measured citation support at 96.4%, with 10/10 completed. Gemini 2.5 Flash ran every agent and Claude Sonnet 4.5 judged. That is one run of ten questions on one model routing, and support means a claim matches the evidence it cites, not that the claim is true. A planner-only candidate measured 0.9389 on 8 of 10, with two queries lost to provider outages, and is reported as below threshold. Cost was not measured.
• A stop holds wherever it lands. Three pre-existing gaps were fixed before this tag (#165).

Desktop builds:

• macOS: Apple Silicon only (…_aarch64.dmg). There is no Intel build; on an Intel Mac, use the Docker version.
• Windows: x64 (.msi).
• Linux: amd64 (.AppImage, .deb).

The builds are unsigned and do not auto-update.

Known: the default Google routing uses Gemini 2.5 models, which Google now limits to accounts that have used them before. If your key is refused, choose other models in Settings → Models. The desktop's first-launch demo fails on a fresh install. The full list is in the changelog. The release gates and their evidence are in internal/rfcs/V3.0-scope-freeze.md §20.

---

Multi-arch container images (linux/amd64, linux/arm64) pushed to GHCR:

• ghcr.io/adityamhaske/multi-agent-research-assistant-api:3.0.0
• ghcr.io/adityamhaske/multi-agent-research-assistant-worker:3.0.0
• ghcr.io/adityamhaske/multi-agent-research-assistant-frontend:3.0.0

latest points at this release.

Deploy the full stack on a free Oracle Cloud Ampere VM — see deploy/README.md.

What's Changed

• chore(deps-dev): bump eslint-config-next from 16.3.3 to 16.3.5 in /frontend by @dependabot[bot] in #138
• docs: close six places where the docs contradicted the code by @adityamhaske in #144
• test(correlation): declare report chat's server log site as session-only by @adityamhaske in #147
• docs(17): record the download page's version selector in §7 by @adityamhaske in #146
• chore(deps): make Dependabot report signal, not churn by @adityamhaske in #145
• chore(deps): update weasyprint requirement from >=69.0 to >=70.0 in /backend by @dependabot[bot] in #149
• chore(deps): update cryptography requirement from >=43.0 to >=50.0.1 in /backend by @dependabot[bot] in #150
• chore(deps): update httpx requirement from >=0.27 to >=0.28.1 in /backend by @dependabot[bot] in #154
• chore(deps): bump the github-actions group with 2 updates by @dependabot[bot] in #156
• chore(deps-dev): bump vitest from 4.1.11 to 5.0.1 in /frontend by @dependabot[bot] in #155
• chore(deps): bump the frontend-minor group in /frontend with 2 updates by @dependabot[bot] in #153
• chore(deps): bump node from 22-bookworm-slim to 25-bookworm-slim in /frontend in the frontend-images group by @dependabot[bot] in #148
• test(desktop): identify the child's payload by shape, not by position by @adityamhaske in #157
• chore(deps): update prometheus-client requirement from >=0.20 to >=0.26.0 in /backend by @dependabot[bot] in #151
• V3 preparation: AP-1–AP-3 and PR-0–PR-6 by @adityamhaske in #158
• V3 PR-7a: dual-version bundle verifier and prompt provenance by @adityamhaske in #159
• V3 PR-7b: capture execution-time prompt provenance by @adityamhaske in #160
• V3 PR-8: custom-spec eval, thresholds and eval isolation by @adityamhaske in #161
• V3 PR-9: prompt-override threat model and bundle v2 documentation by @adityamhaske in #162
• V3 PR-10: per-role agent instructions and what research actually ran under by @adityamhaske in #163
• V3 PR-11: release V3.0.0 with every gate evidenced by @adityamhaske in #164
• V3 release fix: a stop holds wherever it lands (issue #54) by @adityamhaske in #165

Full Changelog: v2.1.0...v3.0.0

Desktop bundles for v3.0.0

These builds are not code-signed. Your OS will warn you on first launch:

OS	What you will see	What to do
macOS	"Apple could not verify this app is free of malware"	System Settings → Privacy & Security → Open Anyway
Windows	SmartScreen warning	More info → Run anyway
Linux	AppImage is not executable	chmod +x *.AppImage (.deb needs nothing)

Verify your download against SHA256SUMS before running it:

sha256sum -c SHA256SUMS --ignore-missing

Research bundles from v3.0.0 onward are format v2, and a verifier from before
v3.0.0 refuses them. Check a bundle with the verify_bundle shipped in this
release; it still verifies every earlier (v1) bundle, unchanged.

V2.1.0 with its gaps stated (#142)

Multi-arch container images (linux/amd64, linux/arm64) pushed to GHCR:

• ghcr.io/adityamhaske/multi-agent-research-assistant-api:2.1.0
• ghcr.io/adityamhaske/multi-agent-research-assistant-worker:2.1.0
• ghcr.io/adityamhaske/multi-agent-research-assistant-frontend:2.1.0

latest points at this release.

Deploy the full stack on a free Oracle Cloud Ampere VM — see deploy/README.md.

What's Changed

• chore(deps-dev): bump @types/react-dom from 19.2.4 to 19.2.5 in /frontend in the react group by @dependabot[bot] in #94
• chore(deps): update langchain-openai requirement from >=1.4.2 to >=1.6.0 in /backend by @dependabot[bot] in #95
• chore(deps-dev): bump @testing-library/react from 16.3.2 to 16.3.3 in /frontend by @dependabot[bot] in #96
• chore(deps): update sqlalchemy requirement from <2.1,>=2.0.51 to >=2.0.52,<2.1 in /backend by @dependabot[bot] in #97
• chore(deps): update langchain-anthropic requirement from >=1.6.1 to >=1.7.0 in /backend by @dependabot[bot] in #99
• chore(deps): update pytest requirement from >=8.3 to >=9.1.1 in /backend by @dependabot[bot] in #101
• chore(deps): update langchain-core requirement from >=1.5.5 to >=1.6.1 in /backend by @dependabot[bot] in #103
• chore(deps-dev): bump @types/node from 26.2.0 to 26.4.0 in /frontend by @dependabot[bot] in #102
• chore(deps-dev): bump @testing-library/user-event from 14.6.4 to 14.6.6 in /frontend by @dependabot[bot] in #100
• chore(deps): update langchain-google-genai requirement from >=2.0 to >=4.3.7 in /backend by @dependabot[bot] in #104
• chore(deps): update ruff requirement from >=0.16.1 to >=0.16.5 in /backend by @dependabot[bot] in #105
• chore(deps): bump @tanstack/react-query from 5.101.4 to 5.102.8 in /frontend by @dependabot[bot] in #106
• chore(deps): update pypdf requirement from >=4.0 to >=6.16.2 in /backend by @dependabot[bot] in #108
• chore(deps-dev): bump @vitejs/plugin-react from 6.0.3 to 6.1.1 in /frontend by @dependabot[bot] in #112
• chore(deps-dev): bump eslint-config-next from 16.3.2 to 16.3.3 in /frontend by @dependabot[bot] in #109
• chore(deps): update langgraph-checkpoint-sqlite requirement from >=3.0 to >=3.1.1 in /backend by @dependabot[bot] in #110
• chore(deps): update lxml requirement from >=6.1.1 to >=6.1.2 in /backend by @dependabot[bot] in #107
• chore(deps): bump next from 16.3.2 to 16.3.3 in /frontend by @dependabot[bot] in #111
• SEO, crawlability and a real sitemap for the Pages site by @adityamhaske in #114
• fix(desktop): stop two Windows-only sidecar crashes + repo cleanup by @adityamhaske in #113
• chore(deps): update tavily-python requirement from >=0.5 to >=0.8.1 in /backend by @dependabot[bot] in #115
• chore(deps-dev): bump @types/react-dom from 19.2.5 to 19.2.7 in /frontend in the react group by @dependabot[bot] in #116
• chore(deps): update pydantic requirement from <3.0,>=2.13.4 to >=2.13.5,<3.0 in /backend by @dependabot[bot] in #117
• chore(deps-dev): bump @playwright/test from 1.62.1 to 1.63.0 in /frontend by @dependabot[bot] in #118
• chore(deps-dev): bump @types/node from 26.4.0 to 26.4.1 in /frontend by @dependabot[bot] in #120
• chore(deps): update langchain-core requirement from >=1.6.1 to >=1.6.2 in /backend by @dependabot[bot] in #121
• chore(deps): update python-multipart requirement from >=0.0.9 to >=0.0.32 in /backend by @dependabot[bot] in #123
• chore(deps): bump next from 16.3.3 to 16.3.4 in /frontend by @dependabot[bot] in #124
• chore(deps-dev): bump browserslist from 4.28.2 to 4.28.9 in /frontend by @dependabot[bot] in #125
• chore(deps): bump sharp from 0.35.3 to 0.35.4 in /frontend by @dependabot[bot] in #126
• chore(deps): bump baseline-browser-mapping from 2.10.38 to 2.11.21 in /frontend by @dependabot[bot] in #127
• chore(deps-dev): bump brace-expansion from 1.1.15 to 1.1.18 in /frontend by @dependabot[bot] in #128
• chore(deps-dev): bump js-yaml from 4.2.0 to 4.3.2 in /frontend by @dependabot[bot] in #129
• V2.1-a foundations: airgap, fake-mode, chat window, retrieval eval, corpus versioning, typed artifacts, observability, migration integrity (A1–A9) by @adityamhaske in #130
• V2.1-a A10: make server and desktop readiness hermetic by @adityamhaske in #131
• chore(deps): update tavily-python requirement from >=0.8.1 to >=0.8.2 in /backend by @dependabot[bot] in #132
• chore(deps): update alembic requirement from >=1.19.1 to >=1.20.0 in /backend by @dependabot[bot] in #133
• chore(deps): update lxml requirement from >=6.1.2 to >=6.1.3 in /backend by @dependabot[bot] in #134
• chore(deps): bump the react group in /frontend with 4 updates by @dependabot[bot] in #135
• chore(deps): update langchain-openai requirement from >=1.6.0 to >=1.6.2 in /backend by @dependabot[bot] in #136
• chore(deps): update psycopg requirement from >=3.3.4 to >=3.3.5 in /backend by @dependabot[bot] in #137
• chore(deps-dev): bump @types/node from 26.4.1 to 26.5.1 in /frontend by @dependabot[bot] in #139
• chore(deps): bump next from 16.3.3 to 16.3.5 in /frontend by @dependabot[bot] in #140
• chore(deps-dev): bump @testing-library/user-event from 14.6.6 to 14.6.7 in /frontend by @dependabot[bot] in #141
• release(v2.1.0): cut V2.1 with its gaps stated by @adityamhaske in #142

Full Changelog: v2.0.2...v2.1.0

Desktop bundles for v2.1.0

These builds are not code-signed. Your OS will warn you on first launch:

OS	What you will see	What to do
macOS	"Apple could not verify this app is free of malware"	System Settings → Privacy & Security → Open Anyway
Windows	SmartScreen warning	More info → Run anyway
Linux	AppImage is not executable	chmod +x *.AppImage (.deb needs nothing)

Verify your download against SHA256SUMS before running it:

sha256sum -c SHA256SUMS --ignore-missing

v2.0.2

Multi-arch container images (linux/amd64, linux/arm64) pushed to GHCR:

• ghcr.io/adityamhaske/multi-agent-research-assistant-api:2.0.2
• ghcr.io/adityamhaske/multi-agent-research-assistant-worker:2.0.2
• ghcr.io/adityamhaske/multi-agent-research-assistant-frontend:2.0.2

latest points at this release.

Deploy the full stack on a free Oracle Cloud Ampere VM — see deploy/README.md.

Full Changelog: v2.0.1...v2.0.2

Desktop bundles for v2.0.2

These builds are not code-signed. Your OS will warn you on first launch:

OS	What you will see	What to do
macOS	"Apple could not verify this app is free of malware"	System Settings → Privacy & Security → Open Anyway
Windows	SmartScreen warning	More info → Run anyway
Linux	AppImage is not executable	chmod +x *.AppImage (.deb needs nothing)

Verify your download against SHA256SUMS before running it:

sha256sum -c SHA256SUMS --ignore-missing

v2.0.1

Desktop bundles for v2.0.1.

These builds are not code-signed. Your OS will warn you on first launch:

OS	What you will see	What to do
macOS	"Apple could not verify this app is free of malware"	System Settings → Privacy & Security → Open Anyway
Windows	SmartScreen warning	More info → Run anyway
Linux	AppImage is not executable	chmod +x *.AppImage (.deb needs nothing)

Verify your download against SHA256SUMS before running it:

sha256sum -c SHA256SUMS --ignore-missing

---

Multi-arch container images (linux/amd64, linux/arm64) pushed to GHCR:

ghcr.io/adityamhaske/multi-agent-research-assistant-api:2.0.1
ghcr.io/adityamhaske/multi-agent-research-assistant-worker:2.0.1
ghcr.io/adityamhaske/multi-agent-research-assistant-frontend:2.0.1

latest points at this release.

Deploy the full stack on a free Oracle Cloud Ampere VM — see deploy/README.md.

v2.0.0

Desktop bundles for v2.0.0.

These builds are not code-signed. Your OS will warn you on first launch:

OS	What you will see	What to do
macOS	"Apple could not verify this app is free of malware"	System Settings → Privacy & Security → Open Anyway
Windows	SmartScreen warning	More info → Run anyway
Linux	AppImage is not executable	chmod +x *.AppImage (.deb needs nothing)

Verify your download against SHA256SUMS before running it:

sha256sum -c SHA256SUMS --ignore-missing

v1.0.2

Desktop bundles for v1.0.2.

These builds are not code-signed. Your OS will warn you on first launch:

OS	What you will see	What to do
macOS	"Apple could not verify this app is free of malware"	System Settings → Privacy & Security → Open Anyway
Windows	SmartScreen warning	More info → Run anyway
Linux	AppImage is not executable	chmod +x *.AppImage (.deb needs nothing)

Verify your download against SHA256SUMS before running it:

sha256sum -c SHA256SUMS --ignore-missing