TODO items commented in code

[YashGarg-4]

Can I pick up the TODO items commented in code? Are they all posted as individual issues?

[eeshsaxena]

Not a maintainer, but I went and counted since I had the repo open, in case it saves you the grep.

There are 33 TODO/FIXME comments under src/ on current main, and none of them reference an issue number. A search for open issues mentioning TODO comes back empty too, so they don't look like they're individually tracked. They read more like notes left in place while writing each provider than a curated backlog.

The other thing worth knowing before you pick one: they're not 33 distinct problems. The single biggest cluster is the same sentence copy-pasted across providers as each one was added, some spelled TODO and some TODOS:

providers/anthropic/chatComplete.ts:28
providers/anyscale/chatComplete.ts:13
providers/azure-ai-inference/chatComplete.ts:9
providers/azure-ai-inference/complete.ts:4
providers/azure-openai/chatComplete.ts:10
providers/azure-openai/complete.ts:5
providers/azure-openai/embed.ts:7

all saying the config doesn't enforce the max token limit for the input. That's one design question about where validation belongs, not seven tasks, and it'd be worth agreeing on the approach in an issue before touching seven files.

The ones that look genuinely self-contained are the specific ones, for example providers/anthropic/complete.ts:75 ("the token calculation is wrong atm") and providers/bedrock/chatComplete.ts:395 (splitting a provider option in two). Those are small enough to just open a PR for.

If it were me I'd open one issue for the max-token-limit cluster to get a maintainer's read on the intended design first, and send the narrow ones straight as PRs. Worth waiting for someone from the team to confirm, though, since they may already have a view on which of these are deliberate.