✅ #7 Pre-Phase-7: per-source LLM routing seam +452 -16 11d ago self by
Archibald312 ↗ analysis ↗ GitHub ## Summary
Archibald312/GordonOSS
A finance-focused fork of Mike, reborn as Gordon - an AI analyst built for deal diligence, spreadsheets, and paid-data research.
This is a hard fork that's no longer trying to be a legal-AI tool. Archibald312 has repositioned the project wholesale toward finance work - M&A diligence, private equity, private credit, leveraged finance, and equity/credit research - and rebranded it from Mike into Gordon, an AI analyst with a name and a niche of its own. The package identity, the persona, and the product framing have all moved in the same direction.
What stands out is how much groundwork came before the features. Archibald312 stood up continuous integration, backend and end-to-end tests, linting, and a privacy guard that controls when documents are routed through AI - the kind of safety net many open-source AI projects skip. On top of that foundation, the fork now treats spreadsheets as first-class material: it ingests Excel and CSV files and has the assistant cite answers down to the exact cell.
The trajectory is clearly toward plugging in outside, paid data feeds. Before turning on those connectors, Archibald312 has laid down a per-source model-routing layer (deciding which AI answers which question) and a tamper-evident audit log that records every AI call and tool run - a record even someone with database access can't quietly rewrite. If you want the depth, the GitHub repo is worth a click.
What's in it
- Gordon, the finance analyst The assistant is purpose-built for M&A diligence, private equity, private credit, leveraged finance, and equity/credit research - not general legal work.
- Spreadsheets as first-class material Ingests Excel and CSV files end-to-end and cites its answers down to the exact cell, so you can trace any claim back to the source.
- Tamper-evident audit trail Every AI call and tool invocation is recorded in a log designed to resist rewriting - even by someone holding database credentials.
- Per-source model routing A switchboard decides which AI model handles which question, laid down ahead of the external data connectors that will rely on it.
- Privacy guard on AI calls Controls when documents actually get routed through AI, drawing a clear line for anyone sending sensitive material through the system.
- A real safety net Continuous integration, backend unit tests, and revived end-to-end suites landed before the features did - guardrails first.
Direction
brandinginfrastructureanalytics
Activity
✅ #6 Phase 6: audit logging for LLM + tool invocations +608 -14 11d ago self by
Archibald312 ↗ analysis ↗ GitHub ## Summary
📝 Archibald312 builds the routing dial before turning it 1 commit 11d ago infrastructurecompliance draft
GordonOSS lands a per-source model-routing layer ahead of the connectors that will need it - and pins the rules with tests before anyone can break them.
✅ #5 Phase 4: Domain swap (legal → finance) + Mike → Gordon rename +650 -435 11d ago self by
Archibald312 ↗ analysis ↗ GitHub ## Summary
📝 Archibald312 builds a tamper-evident paper trail for every AI move 1 commit 11d ago compliancesecurity draft
GordonOSS now records every AI call and tool invocation in a log that even an attacker with database credentials cannot rewrite.
A finance-focused fork of Mike now ingests Excel and CSV files end-to-end, with the AI citing answers down to the exact cell.
📝 Archibald312 turns Mike into Gordon, a finance-AI assistant 1 commit 11d ago brandingsummarization draft
The fork pivots wholesale from legal work to M&A diligence, private equity, private credit, leveraged finance, and equity/credit research.
✅ #4 Free-tier LLM dev-mode fixes + re-enable e2e product-flow specs +354 -201 11d ago self by
Archibald312 ↗ analysis ↗ GitHub ## Summary
📝 Re-enable product-flow e2e specs (projects, documents, chat, tabular) (no public page) 1 commit 11d ago not yet rewritten
✅ #3 fix: allow free-tier LLM in dev mode, improve error visibility +22 -5 12d ago self by
Archibald312 ↗ analysis ↗ GitHub ## Summary
✅ #2 refactor(backend): split chatTools.ts into per-tool registry +2,576 -2,344 12d ago self by
Archibald312 ↗ analysis ↗ GitHub ## Summary
✅ #1 Add CI pipeline, test suite, and finance-fork infrastructure +7,449 -588 12d ago self by
Archibald312 ↗ analysis ↗ GitHub ## Summary
🟢 #123 chore: rename packages from mike to GordonOSS +7,449 -588 12d ago by
Archibald312 ↗ analysis ↗ GitHub ## Summary
📝 Free-tier LLM guard usability fixes (no public page) 1 commit 12d ago not yet rewritten
📝 chatTools.ts refactor into per-tool registry (no public page) 1 commit 12d ago not yet rewritten
📝 Archibald312 puts a real safety net under the codebase 1 commit 12d ago infrastructuresecurity draft
Before adding features, this fork built the testing and guardrails most open-source legal-AI projects skip.
📝 Rename packages from mike to GordonOSS to match the fork (no public page) 0 commits — not yet rewritten
A wave of engineering plumbing makes GordonOSS ready to live on its own - and quietly draws a privacy line that matters to anyone routing client documents through AI.
A 3,000-line monolith that wired the chat assistant to everything it can do gets carved into a clean registry, so future capabilities slot in without surgery.
📝 Unblock free-tier LLM calls and surface real errors (no public page) 0 commits — not yet rewritten
Show 4 more
📝 Archibald312 unblocks the free tier and turns the test lights back on 0 commits — infrastructure draft
A two-part fix that makes the fork easier to try and harder to break.
A single phase repositions the fork from legal AI to finance AI - and finishes the brand rename in the same sweep.
📝 Archibald312 wires GordonOSS for paid-data accountability 0 commits — complianceinfrastructure draft
Before letting third-party data feeds plug in, the fork is building a tamper-proof record of every AI call and every tool the system runs.
📝 Archibald312 builds the switchboard before the switches 0 commits — infrastructurecompliance draft
GordonOSS shelved local-model support to first lay down the wiring that decides which AI answers which question.
Threads of work (detailed view)
Archibald312 builds the routing dial before turning it
GordonOSS lands a per-source model-routing layer ahead of the connectors that will need it - and pins the rules with tests before anyone can break them.
Archibald312 builds a tamper-evident paper trail for every AI move
GordonOSS now records every AI call and tool invocation in a log that even an attacker with database credentials cannot rewrite.
Archibald312 makes spreadsheets first-class in GordonOSS
A finance-focused fork of Mike now ingests Excel and CSV files end-to-end, with the AI citing answers down to the exact cell.
Archibald312 turns Mike into Gordon, a finance-AI assistant
The fork pivots wholesale from legal work to M&A diligence, private equity, private credit, leveraged finance, and equity/credit research.
Archibald312 puts a real safety net under the codebase
Before adding features, this fork built the testing and guardrails most open-source legal-AI projects skip.
Archibald312 builds the switchboard before the switches
GordonOSS shelved local-model support to first lay down the wiring that decides which AI answers which question.
Archibald312 wires GordonOSS for paid-data accountability
Before letting third-party data feeds plug in, the fork is building a tamper-proof record of every AI call and every tool the system runs.
Archibald312 turns Mike into a finance analyst named Gordon
A single phase repositions the fork from legal AI to finance AI - and finishes the brand rename in the same sweep.
Archibald312 unblocks the free tier and turns the test lights back on
A two-part fix that makes the fork easier to try and harder to break.
Archibald312 makes adding new AI tools a one-file job
A 3,000-line monolith that wired the chat assistant to everything it can do gets carved into a clean registry, so future capabilities slot in without surgery.
Archibald312 lays the foundations for a hard fork
A wave of engineering plumbing makes GordonOSS ready to live on its own - and quietly draws a privacy line that matters to anyone routing client documents through AI.
Pull requests (detailed view)
🟢 Open (1)
by
Archibald312 · opened 13d ago ✅ Merged (7)
by
Archibald312 · opened 11d ago · merged 11d ago by Archibald312 by
Archibald312 · opened 11d ago · merged 11d ago by Archibald312 by
Archibald312 · opened 11d ago · merged 11d ago by Archibald312 by
Archibald312 · opened 12d ago · merged 11d ago by Archibald312 by
Archibald312 · opened 12d ago · merged 12d ago by Archibald312 by
Archibald312 · opened 12d ago · merged 12d ago by Archibald312 by
Archibald312 · opened 12d ago · merged 12d ago by Archibald312