mrdulasolutions makes Mike's AI provider a real choice

This fork lets a legal team point Mike at compatible AI services without locking the product to one provider.

integrationsecurity

mrdulasolutions turns model selection into an operator decision, with the chosen service and model options appearing in Mike's model picker rather than living as a hidden technical assumption.

Compatible AI endpoints: connect Mike to other OpenAI-compatible services, including self-hosted options.
Flexible model catalogue: present approved high, mid and lower-cost models with clear labels for users.
Reliable streaming: improve how live AI responses finish and release resources.
Safer configuration: block risky endpoint addresses, avoid exposing credentials, limit oversized model lists and time out stalled requests.

The follow-up hardening matters: changing an AI provider connection can create security and reliability risks, and this fork treats those as part of the feature rather than cleanup for later.

So what Legal-ops leads and product teams weighing provider flexibility should inspect this fork if they need more control over cost, hosting or model choice.

Spotted something wrong? Or know the PR text has fresher detail than the writeup above?

Commits in this thread

2 commits from mrdulasolutions/mike, oldest first. Source extracted verbatim from the harvested git log.

SHA	Subject	Author	Date
`9f370da0`	feat: support any OpenAI-compatible LLM endpoint	MRDula	2026-07-26	↗ GitHub
commit body Allow OPENAI_BASE_URL plus chat/completions vs Responses API mode so Mike can call Cerebras, Bedrock Mantle, and other OpenAI-compatible servers. Model picker loads from GET /config/models with env-driven OPENAI_MODELS and optional OPENAI_COMPAT_ANY_MODEL.
`93ad5a68`	harden OpenAI-compatible adapter for production	MRDula	2026-07-26	↗ GitHub
commit body Validate OPENAI_BASE_URL (https, no embedded credentials), add request timeouts, truncate upstream error bodies, flush trailing SSE frames, release stream readers, sanitize model ids, and fix the frontend model picker accept logic. Public /config/models only exposes host + mode.

SHA

Subject

Author

Date

9f370da0

feat: support any OpenAI-compatible LLM endpoint

MRDula

2026-07-26

↗ GitHub

commit body

Allow OPENAI_BASE_URL plus chat/completions vs Responses API mode so
Mike can call Cerebras, Bedrock Mantle, and other OpenAI-compatible
servers. Model picker loads from GET /config/models with env-driven
OPENAI_MODELS and optional OPENAI_COMPAT_ANY_MODEL.

93ad5a68

harden OpenAI-compatible adapter for production

MRDula

2026-07-26

↗ GitHub

commit body

Validate OPENAI_BASE_URL (https, no embedded credentials), add request
timeouts, truncate upstream error bodies, flush trailing SSE frames,
release stream readers, sanitize model ids, and fix the frontend model
picker accept logic. Public /config/models only exposes host + mode.

Capture this thread into my fork

Download a single Markdown prompt that tells Claude how to port every commit above into your working tree — adapting paths and structure to match your repo. Run it via claude -p < capture-thread-986.md from inside the repo you want the changes in.