Migrated from the archived repo: Database-Tycoon/localhost-stack#14 (opened 2026-04-01). Content below is the original issue verbatim; status notes at the end.
Background
The current tycoon ai stack is hard-wired to LM Studio's local server at localhost:1234. LM Studio is a heavy desktop app — it's fine for development on a high-end machine but is not a realistic dependency for general-purpose users.
Research findings
- OpenAI Codex CLI — cloud-only, conflicts with local-first design. Not worth pursuing.
- OpenAI gpt-oss-20b — Apache 2.0 open-weight model, 16GB RAM, available on Ollama. Viable locally, but adds Ollama as a dependency.
- Ollama — lighter than LM Studio, broader hardware support, CLI-first, no GUI. But still a separate install.
Goal
Find the single shortest path to running a capable model for the specific tasks tycoon needs:
TestFixer — fix a failing dbt testColumnDocumenter — generate schema.yml descriptionsStagingImprover — refactor a staging model
These are focused, single-turn tasks with small context windows. They do not need a general-purpose chat model or a large context.
Design questions to answer
- What is the lightest runtime that can serve a small model locally without a GUI dependency? (Ollama, llama.cpp server, mlx, transformers pipeline?)
- Is there a model small enough to run on CPU-only hardware that is still useful for these tasks? (e.g. Qwen2.5-Coder-1.5B, Phi-3-mini)
- Should
tycoon ai ship with a recommended model + one-line install command, rather than requiring the user to set up LM Studio separately?
- Can we reduce
tycoon.ai.client to a single, minimal HTTP call with no backend abstraction layer?
Out of scope
- Multiple LLM backend options / provider abstraction layer
- Cloud API fallbacks
- OpenAI Codex CLI integration
Status notes at migration (2026-08-09)
Mirrored in Jira: PTC-115 (Epic: PTC-5, AI / agent refinements). Dev work continues here.
Background
The current
tycoon aistack is hard-wired to LM Studio's local server atlocalhost:1234. LM Studio is a heavy desktop app — it's fine for development on a high-end machine but is not a realistic dependency for general-purpose users.Research findings
Goal
Find the single shortest path to running a capable model for the specific tasks tycoon needs:
TestFixer— fix a failing dbt testColumnDocumenter— generate schema.yml descriptionsStagingImprover— refactor a staging modelThese are focused, single-turn tasks with small context windows. They do not need a general-purpose chat model or a large context.
Design questions to answer
tycoon aiship with a recommended model + one-line install command, rather than requiring the user to set up LM Studio separately?tycoon.ai.clientto a single, minimal HTTP call with no backend abstraction layer?Out of scope
Status notes at migration (2026-08-09)
feat/okf-knowledge-bundle("implement lightweight LLM backend (Ollama)"), written against the pre-rewrite codebase — treat as reference, not a mergeable base.askextra and its dead code were dropped in Drop server/dagster/ask extras and their dead code (rewrite deletion milestone) #139, and tycoon ask has no headless/CLI interface — forced to bypass Nao to use the local LLM from the terminal #76 tracks a headless CLI path to the local LLM. Retriage against the current AI story before picking up.Mirrored in Jira: PTC-115 (Epic: PTC-5, AI / agent refinements). Dev work continues here.