Pin pydantic to v2 so install doesn't crash on import #139

Merged
TanmayDoesAI merged 1 commit from fix/setup-reliability-and-hardware-support into main 2026-06-01 07:23:50 +02:00
TanmayDoesAI commented 2026-05-31 21:55:11 +02:00 (Migrated from github.com)

Rebased on current main and trimmed this down. When I went through it against main, most of what I'd bundled here is already handled:

  • the calendar f-string fix is already in (same change)
  • the host.docker.internal / Ollama note is already in the README
  • the clone placeholders are already replaced with the real URL

So the only thing still missing is the pydantic pin, which is what this PR is now. Unpinned, pip can resolve v1 which has no pydantic-core and the app dies on import.

Dropped the Intel GPU detection from here. jaitaiwan's point about integrated Arc/Xe (the 140V etc.) using unified memory is right, my version hardcoded that to false. I'll redo it as its own PR with the unified-memory case handled properly.

Rebased on current main and trimmed this down. When I went through it against main, most of what I'd bundled here is already handled: - the calendar f-string fix is already in (same change) - the host.docker.internal / Ollama note is already in the README - the clone placeholders are already replaced with the real URL So the only thing still missing is the pydantic pin, which is what this PR is now. Unpinned, pip can resolve v1 which has no pydantic-core and the app dies on import. Dropped the Intel GPU detection from here. jaitaiwan's point about integrated Arc/Xe (the 140V etc.) using unified memory is right, my version hardcoded that to false. I'll redo it as its own PR with the unified-memory case handled properly.
jaitaiwan commented 2026-06-01 03:38:13 +02:00 (Migrated from github.com)

This PR assumes that intel arc GPUs aren't integrated with unified memory. If this was run on a device like a Yoga Slim 7i which has a built-in Intel Arc 140V then it uses unified memory.

This PR assumes that intel arc GPUs aren't integrated with unified memory. If this was run on a device like a Yoga Slim 7i which has a built-in Intel Arc 140V then it uses unified memory.
pewdiepie-archdaemon commented 2026-06-01 06:05:48 +02:00 (Migrated from github.com)

This bundles several useful fixes, but it looks stale/too mixed to merge as-is.

A couple of notes from checking against current main:

  • The README already has a host.docker.internal / Ollama note now, so that part may duplicate or conflict after rebase.
  • The pydantic and Python <3.12 f-string fixes are small and probably worth splitting into their own focused PR.
  • The Intel GPU detection should address the unified-memory point raised above before merge. Returning unified_memory: false for all Intel vendor 0x8086 devices can misclassify Arc/Xe integrated GPUs.

Recommendation: rebase and split this into install/docs fixes vs Intel hardware detection. The install fixes are likely easy to review; the Intel path needs a bit more hardware-specific care.

This bundles several useful fixes, but it looks stale/too mixed to merge as-is. A couple of notes from checking against current main: - The README already has a host.docker.internal / Ollama note now, so that part may duplicate or conflict after rebase. - The pydantic and Python <3.12 f-string fixes are small and probably worth splitting into their own focused PR. - The Intel GPU detection should address the unified-memory point raised above before merge. Returning unified_memory: false for all Intel vendor 0x8086 devices can misclassify Arc/Xe integrated GPUs. Recommendation: rebase and split this into install/docs fixes vs Intel hardware detection. The install fixes are likely easy to review; the Intel path needs a bit more hardware-specific care.
TanmayDoesAI commented 2026-06-01 07:21:03 +02:00 (Migrated from github.com)

@pewdiepie-archdaemon done, rebased on main and split it. Turned out the calendar fix, the docker note, and the clone placeholders are all already on main, so this PR is just the pydantic pin now. Small and easy to review like you said.

@jaitaiwan good catch, you're right about the integrated parts like the 140V sharing memory with the CPU. Let me look into this properly, I'll do a bit of research on it and put up a separate PR for the Intel detection.

@pewdiepie-archdaemon done, rebased on main and split it. Turned out the calendar fix, the docker note, and the clone placeholders are all already on main, so this PR is just the pydantic pin now. Small and easy to review like you said. @jaitaiwan good catch, you're right about the integrated parts like the 140V sharing memory with the CPU. Let me look into this properly, I'll do a bit of research on it and put up a separate PR for the Intel detection.
Sign in to join this conversation.
No description provided.