feat: add Avian as an LLM provider

[avianion]

Summary

• Adds Avian as a first-class LLM provider for PrivateGPT
• Avian exposes an OpenAI-compatible API (https://api.avian.io/v1) with access to frontier models at competitive pricing
• Reuses the existing llama-index-llms-openai / llama-index-embeddings-openai integrations, so no new dependencies are introduced

Changes

File	Change
private_gpt/settings/settings.py	Add "avian" to LLMSettings.mode and EmbeddingSettings.mode Literals; add AvianSettings config model
private_gpt/components/llm/llm_component.py	Add "avian" case using llama_index.llms.openai.OpenAI pointed at api.avian.io/v1
private_gpt/components/embedding/embedding_component.py	Add "avian" case using llama_index.embeddings.openai.OpenAIEmbedding
pyproject.toml	Add llms-avian and embeddings-avian extras (pointing to existing openai packages)
settings-avian.yaml	New profile for quick Avian setup
settings.yaml	Add avian: defaults section

Available Models

Model	Context	Max Output	Input / Output (per 1M tokens)
deepseek/deepseek-v3.2	164K	65K	$0.26 / $0.38
moonshotai/kimi-k2.5	131K	8K	$0.45 / $2.20
z-ai/glm-5	131K	16K	$0.30 / $2.55
minimax/minimax-m2.5	1M	1M	$0.30 / $1.10

Usage

# Install with Avian support
poetry install --extras "llms-avian embeddings-avian vector-stores-qdrant ui"

# Set API key
export AVIAN_API_KEY=your-key-here

# Run with Avian profile
PGPT_PROFILES=avian make run

Design Decisions

• Reuses OpenAI llama-index integration: Since Avian is fully OpenAI-compatible (chat completions, streaming, function calling), this PR reuses the existing llama-index-llms-openai package rather than adding a new dependency. This follows the same approach used by the openai mode but with Avian-specific defaults (API base URL, model names, settings).
• Separate AvianSettings config: Gives users a clean, dedicated configuration section instead of requiring them to manually override the openai settings block with Avian's API base and key.
• No breaking changes: All existing modes and settings are unchanged. The new avian sections in settings.yaml use env var defaults (${AVIAN_API_KEY:}) so they don't interfere with existing deployments.

Test Plan

• Verify existing tests pass (no breaking changes to settings schema)
• Test PGPT_PROFILES=avian loads correct settings
• Test chat completion with deepseek/deepseek-v3.2 model
• Test streaming responses work correctly

cc @imartinez @jaluma @lopagela for review

[avianion]

Hey @jaluma @imartinez, would love your review on this when you get a chance. Happy to address any feedback!

[avianion]

Friendly follow-up — this PR is still active and ready for review. Would appreciate a look when you get a chance! cc @jaluma @imartinez

[avianion]

Friendly follow-up — this PR is still active and ready for review. All feedback has been addressed. Would appreciate a look when you get a chance! cc @jaluma @imartinez

[avianion]

Hey @imartinez @jaluma — friendly follow-up on this PR. Avian is an OpenAI-compatible inference provider that's already live and powering apps like ISEKAI ZERO. This is a lightweight integration (standard OpenAI-compatible endpoint) and we're happy to address any feedback or make adjustments. Would love to get this merged if you have a moment to review. Thanks!

[imartinez]

Thanks for the contribution. I think your Usage example is not right:
Current:
poetry install --extras "llms-avian embeddings-huggingface vector-stores-qdrant ui"
Expected:
poetry install --extras "llms-avian embeddings-avian vector-stores-qdrant ui"

[imartinez]

Please review the comments, as I feel you missed specifically one change in the settings-avian.yaml file

[imartinez]

You may want to make this configurable via env var like ${AVIAN_ENDPOINT:}

[imartinez]

Shouldn't it be avian instead of huggingface?

[avianion]

Addressed all review feedback in commit 761fecc:

• settings-avian.yaml: switched embedding.mode from huggingface to avian (removed the unused huggingface section)
• settings.yaml / settings-avian.yaml: made api_base configurable via env var (${AVIAN_API_BASE:https://api.avian.io/v1})
• PR description: fixed poetry install example to use embeddings-avian instead of embeddings-huggingface

[avianion]

All review feedback has been addressed. Could you please re-review when you get a chance?

[avianion]

Both review comments have been addressed in commit 761fecc:

1. settings.yaml — api_base is now configurable via env var: ${AVIAN_API_BASE:https://api.avian.io/v1} (with a sensible default so existing setups aren't broken)

2. settings-avian.yaml — embedding.mode is now avian (was huggingface). The unused huggingface: section was also removed from that file.

The poetry install example in the PR description also uses embeddings-avian (not embeddings-huggingface).

Happy to rename AVIAN_API_BASE → AVIAN_ENDPOINT if you prefer that naming — just let me know.