docs: add adalat-ai/koyal-indic-600m-1.0 to featured community checkp…

[janaab11]

What does this PR do ?

Submitting koyal-indic-600m-1.0 for the Featured Community Checkpoints page: an open ASR model for Indian languages that emits document-ready text (punctuation and formatted numerals), released by Adalat AI.

Release post: https://www.adalat.ai/blog/tech/koyal-model-release

Collection: ASR (docs)

koyal-indic-600m-1.0 (Nemotron fine-tune)

• Model: https://huggingface.co/adalat-ai/koyal-indic-600m-1.0
• NeMo base checkpoint: nvidia/nemotron-3.5-asr-streaming-0.6b (encoder, kept cache-aware; new prompt-conditioned RNN-T decoder and 6k unified vocabulary for the four languages)
• Task: multilingual ASR with rich-orthography output; dual-mode — one checkpoint serves both streaming (0 ms to 1040 ms lookahead) and full-context offline decoding, selected at inference
• Languages: Hindi, Kannada, Malayalam, Telugu. To our knowledge the first open cache-aware streaming ASR covering Kannada, Malayalam, and Telugu (cf. Extending NVIDIA Nemotron-3.5-ASR-Streaming to Unsupported Languages and Scripts #15793 on base-tokenizer script coverage)
• Evaluation: WER_SCRIBE on FLEURS-RO (rich-orthography FLEURS re-annotation, scribe-eval): offline 12.33 (hi) / 12.75 (kn) / 18.58 (ml) / 20.65 (te); zero-lookahead streaming stays within 2.2 points of offline in every language. Per-operating-point tables and jiwer WER on the model card
• Framework: NeMo (≥ 3.0.0; uses the upstream prompt-conditioned RNN-T class)

Changelog

• docs/source/asr/featured_community_checkpoints.rst: add one table row for adalat-ai/koyal-indic-600m-1.0.

Before your PR is "Ready for review"

Pre checks:

• Make sure you read and followed Contributor guidelines
• Did you write any new necessary tests? — n/a, documentation only
• Did you add or update any necessary documentation?
• Does the PR affect components that are optional to install? — no

PR Type:

• New Feature
• Bugfix
• Documentation

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[janaab11]

tagging @sagarsdesai for visibility!