⛔ #378 fix(models): stop silently defaulting to Gemini outside Google-OAuth deployments +413 -105 1mo ago by
duncanmcqueen ↗ analysis ↗ GitHub ## Summary
duncanmcqueen/mike
A legal-ops-oriented Mike fork bringing case retrieval, flexible AI choices, and more reliable document handling into one workspace.
duncanmcqueen is shaping Mike into a legal-ops workbench: a chat-led place to work with case materials, legal documents, integrations and a wider choice of AI models.
The direction is practical rather than cosmetic. It adds DingDuff case retrieval to conversations, makes legal PDF text and generated Word documents behave more like lawyers expect, and gives deployments and users greater control over their AI providers. If the approach fits your practice, GitHub has the detail.
What's in it
- DingDuff case retrieval Bring case materials and case-specific documents into an assistant conversation through a user-configured connection.
- Flexible AI model choice Use self-hosted or third-party OpenAI-compatible models, alongside additional subscription-backed model routes.
- Model committees Let several models address the same legal question, then have a chosen chair produce one considered response.
- Legal-document fidelity Preserve more of a PDF's layout during extraction and keep Word section numbering off unless it is wanted.
- More dependable chat Make empty AI responses visible and actionable instead of leaving the conversation looking stuck.
- Personal workspace settings Offer a saved, opt-in dark mode that follows users across sessions.
Direction
integrationworkflowinfrastructure
Activity
✅ #379 fix(chat): surface empty upstream model completions instead of silence +42 1mo ago by
duncanmcqueen ↗ analysis ↗ GitHub ## Summary
## Summary
🟢 #339 feat(llm): add model registry and OpenAI-compatible provider foundation +1,818 1mo ago by
duncanmcqueen ↗ analysis ↗ GitHub ## Summary
## Summary
✅ #337 fix(docx): avoid numbering generated letter paragraphs +220 -29 1mo ago by
duncanmcqueen ↗ analysis ↗ GitHub ## Summary
Users can now choose a darker workspace and have that choice follow them when they return.
📝 duncanmcqueen opens Mike to OpenCode Go's open-model catalogue 4 commits 1mo ago integrationworkflow draft
Mike users can now choose from another subscription-backed route to capable open coding models, without loosening control of who can use them.
🟢 #331 feat(mcp): DingDuff connector support - OAuth redirect discovery and chat case retrieval +263 -56 1mo ago by
duncanmcqueen ↗ analysis ↗ GitHub ## Summary
Generated letters now stay readable by default, with clause numbering reserved for documents that actually need it.
⛔ #364 feat: add Synthetic as a fourth model router +1,619 -223 1mo ago by
duncanmcqueen ↗ analysis ↗ GitHub ## Summary
✅ #340 feat: add OpenCode Go as a third model router +1,512 -275 1mo ago by
duncanmcqueen ↗ analysis ↗ GitHub ## Summary
📝 duncanmcqueen wants Mike to add Synthetic to its model-router bench 1 commit 1mo ago integrationworkflow draft
The proposal would let users choose another subscription gateway for open-weights AI models without making the product's model selection more fragmented.
🟢 #336 fix(pdf): preserve visual layout when extracting PDF text +220 -4 2mo ago by
duncanmcqueen ↗ analysis ↗ GitHub ## Problem
A major sync brought new upstream work into the fork, then required seven fixes to restore the fork's own legal-product features.
📝 duncanmcqueen closes a silent gap in large trademark portfolio searches 1 commit 2mo ago searchintegration draft
The fork now preserves complete owner portfolios instead of quietly losing records when a search returns a lot of marks.
📝 CI lockfile/schema drift and coverage-ratchet cleanup 3 commits 2mo ago minor change
This fork reaches beyond chat to connect legal work, documents and AI choices in one place.
This fork makes DingDuff's case materials available inside an assistant conversation, while keeping the connection's sign-in path tightly controlled.
📝 duncanmcqueen is giving legal PDFs their structure back 0 commits — knowledge-managementsearch draft
A proposed overhaul would make extracted legal text more faithful before it reaches chat, quotes, and document workflows.
Show 4 more
The fork makes it far easier to use locally run or third-party AI models without rebuilding the product.
📝 duncanmcqueen wants legal AI answers to come from a committee, not a solo act 0 commits — workflowintegration draft
This fork turns useful model disagreement into one considered answer, without asking users to run the same prompt repeatedly.
This fork makes AI model choice follow what a user or deployment has actually configured, rather than quietly falling back to Gemini.
A completed AI request that says nothing will now tell the user what happened instead of looking stuck.
Threads of work (detailed view)
duncanmcqueen gives Mike a deliberate dark mode
Users can now choose a darker workspace and have that choice follow them when they return.
duncanmcqueen opens Mike to OpenCode Go's open-model catalogue
Mike users can now choose from another subscription-backed route to capable open coding models, without loosening control of who can use them.
duncanmcqueen stops Word from inventing legal numbering
Generated letters now stay readable by default, with clause numbering reserved for documents that actually need it.
duncanmcqueen wants Mike to add Synthetic to its model-router bench
The proposal would let users choose another subscription gateway for open-weights AI models without making the product's model selection more fragmented.
duncanmcqueen repairs a costly upstream sync
A major sync brought new upstream work into the fork, then required seven fixes to restore the fork's own legal-product features.
duncanmcqueen closes a silent gap in large trademark portfolio searches
The fork now preserves complete owner portfolios instead of quietly losing records when a search returns a lot of marks.
duncanmcqueen turns Mike into a legal-ops workbench
This fork reaches beyond chat to connect legal work, documents and AI choices in one place.
duncanmcqueen makes silent chat failures actionable
A completed AI request that says nothing will now tell the user what happened instead of looking stuck.
duncanmcqueen stops Gemini becoming the invisible default
This fork makes AI model choice follow what a user or deployment has actually configured, rather than quietly falling back to Gemini.
duncanmcqueen wants legal AI answers to come from a committee, not a solo act
This fork turns useful model disagreement into one considered answer, without asking users to run the same prompt repeatedly.
duncanmcqueen opens Mike to self-hosted AI models
The fork makes it far easier to use locally run or third-party AI models without rebuilding the product.
duncanmcqueen is giving legal PDFs their structure back
A proposed overhaul would make extracted legal text more faithful before it reaches chat, quotes, and document workflows.
duncanmcqueen brings DingDuff case law into the chat
This fork makes DingDuff's case materials available inside an assistant conversation, while keeping the connection's sign-in path tightly controlled.
Pull requests (detailed view)
🟢 Open (4)
by
duncanmcqueen · opened 2mo ago by
duncanmcqueen · opened 2mo ago 🟢 open ·
#331 feat(mcp): DingDuff connector support - OAuth redirect discovery and chat case retrieval by
duncanmcqueen · opened 2mo ago by
duncanmcqueen · opened 2mo ago ✅ Merged (4)
by
duncanmcqueen · opened 1mo ago · merged 1mo ago by willchen96 by
duncanmcqueen · opened 2mo ago · merged 1mo ago by willchen96 by
duncanmcqueen · opened 2mo ago · merged 1mo ago by willchen96 by
duncanmcqueen · opened 2mo ago · merged 1mo ago by willchen96 ⛔ Closed without merge (2)
by
duncanmcqueen · opened 1mo ago · closed 1mo ago by
duncanmcqueen · opened 1mo ago · closed 1mo ago