Spaces:
Sleeping
Sleeping
Commit History
-
feat: add projects/tinybard/modal_inference.py and update configs
Hermes BotUnverified signature523e468 -
debug: add internet diagnostics to test httpbin
Hermes BotUnverified signaturec326da8 -
feat: add support for custom OpenAI-compatible endpoints in config UI
Hermes BotUnverified signature5941741 -
debug: add env diagnostics to print all env keys
Hermes BotUnverified signaturef525e1a -
fix: target api.huggingface.co instead of api-inference to bypass DNS blocks
Hermes BotUnverified signature445e1fc -
fix: use httpx.Client(trust_env=True) to route through Space environment proxies
Hermes BotUnverified signature167678e -
fix: target huggingface.co domain for serverless API to bypass DNS blocks
Hermes BotUnverified signature1f55836 -
fix: pass user overrides token and model to inference_generate
Hermes BotUnverified signature1a5d3d0 -
debug: add startup diagnostics to test various models and endpoints
Hermes BotUnverified signature16dffaa -
fix: switch to native text_generation for free serverless API
Hermes BotUnverified signatureBy using text_generation instead of chat_completion, we bypass the paid router.huggingface.co endpoint entirely and query the free serverless model endpoint directly via a plain dialogue prompt.2a059de -
fix: resolve base_url and model conflict in InferenceClient
Hermes BotUnverified signatureOmit the model parameter from the constructor and pass the direct model-specific serverless API endpoint as base_url. This forces the use of the free serverless Inference API and resolves the conflict.f42c3c8 -
fix: route chat completions to free serverless Inference API
Hermes BotUnverified signatureBy default, recent huggingface_hub versions route InferenceClient.chat_completion() to router.huggingface.co, which requires paid providers. Explicitly passing base_url='https://api-inference.huggingface.co' forces it to use the free serverless inference endpoint.f438be9 -
fix: lock default model to Qwen/Qwen2.5-1.5B-Instruct
Hermes BotUnverified signatureRouter rejects newer models on this account; keep working free-tier default. Users can override via /api/config modal if they have access to other models.46f8d38 -
fix: revert default model to Qwen/Qwen2.5-1.5B-Instruct
Hermes BotUnverified signatureQwen 2.5 7B requires paid Inference Providers on this account. Revert to 1.5B free-tier model; users can override via config modal.780d3c3 -
fix: switch default model to Qwen/Qwen2.5-7B-Instruct
Hermes BotUnverified signature- Gemma 2 9B unsupported by HF Inference router - Qwen 2.5 7B is supported and strong for storytelling46f4d3a -
fix: handle None genre from Gradio API input
Hermes BotUnverified signature- Default genre to 'fantasy' when Gradio passes None - Fix generate_procedural_step None check - Resolve AttributeError: 'NoneType' object has no attribute 'lower'c4c7f89 -
feat: upgrade to google/gemma-2-9b-it
Hermes BotUnverified signature- 9B parameter model for better storytelling quality - Still under 32B constraint - 3-call generation: story + choices + health commentaryb58e783 -
feat: 3-call LLM generation per turn with funny health commentary
Hermes BotUnverified signature- Story beat generated independently from choices and health effect - Model decides health delta (-15/0/+10) with funny one-liner - Overheal (>=100) and death (<=0) get unique model-generated humor - Infinite gameplay until health hits 0 or reaches 100+ - All 3 generations use Qwen2.5-1.5B-Instruct via HF Inference API5e96ef1 -
feat: LLM-driven infinite adventure loop
Hermes BotUnverified signature- Single generate_llm_turn() call produces story + choices + health delta - Infinite gameplay until health <= 0, no fixed step cap - Procedural fallback preserved during cooldown/failure140c4d5 -
Upload folder using huggingface_hub
Verified signatureb930474 -
initial commit
Verified signaturecdb7ee9
