Spaces:
Sleeping
Sleeping
Commit History
-
Fix Space config: merge duplicate README frontmatter blocks
The hackathon tags were added as a second YAML frontmatter block, so HF parsed only the first (tags-only) block and reported 'Missing SDK in configuration'. Merge tags into the single config block. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>d1e8e99 -
Merge pull request #3 from peteryej/codex/auto-serve-local-llama
Unverified signature[codex] auto-serve local llama models22d1096 -
auto-serve local llama models
ff730f4 -
Merge pull request #2 from peteryej/codex/update-hf_model-to-nvidia-nemotron-3-nano-30b-a3b-bf16
Unverified signatureSet default HF_MODEL to nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF168393072 -
Update Hugging Face model
c50498f -
Add X post link to README
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>f58b0ea -
Add LOCAL_SERVING option to run against a local llama.cpp server
- app.py: LOCAL_SERVING (default False) routes the inference client to a local llama.cpp OpenAI-compatible endpoint (LOCAL_LLM_BASE_URL / LOCAL_LLM_MODEL) instead of the hosted HF Inference API; no HF token needed in that mode. Startup log reports the active backend. - README: document local serving, including the required --reasoning-budget 0 for Gemma 4 (otherwise it stays in thinking mode and returns empty content). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>4004f45 -
Rename Ennui to Boredom, refresh README, pin Space to Gradio 6.16
- emotions.py: change Ennui's display name to Boredom (key unchanged) - README: document clickable emotion chips, .env loading, DEV hot reload, current gemma model + instruct-vs-reasoning note, HF_TOKEN startup log, 100-turn context, FastAPI mount; add example prompts - README frontmatter: bump sdk_version 6.5.1 -> 6.16.0 to match local Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>6a7944a -
Log whether HF_TOKEN is present at startup (boolean only)
Prints '[inside-out] startup: HF_TOKEN present: True/False | model: ...' at import time so the Space logs confirm whether the secret reached the running app. Never logs the secret value. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>94ab97d -
Use gemma instruct model that returns content directly
Qwen3.6-27B is a reasoning model: it returns its answer in message.reasoning and leaves message.content empty, so every reply fell back to canned lines (looked like demo/no-token mode). Switch to google/gemma-4-26B-A4B-it which returns content directly. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>53b0662
-
Fix chat styling, add clickable emotions, and dev ergonomics
UI: - Apply custom CSS/theme via mount_gradio_app (Gradio 6 reads them there, not on gr.Blocks), so the grey message panels actually disappear - Hide the chatbot's top-right Clear/trash button - Turn the emotion legend into clickable chip-buttons: tapping an emotion makes it chime in directly on the conversation so far - Move "Start fresh" above the conversation box Agents: - Feed readable, name-tagged transcripts to the model instead of raw HTML - Widen history context window from 8 to 100 turns - Fall back to canned lines (not the string "None") on empty model replies Dev: - Load .env via python-dotenv if present - DEV=1 enables uvicorn hot reload Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>d7d04f6 -
Disable Gradio SSR mode to fix HF Space port conflict
Gradio 6.x SSR mode spawns a separate Node.js server on port 7861, which conflicts across HF Space restarts and causes a crash loop (Errno 98 address already in use). Disabling SSR removes the Node sidecar and resolves the bind failure. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>daa09b4
-
Merge branch 'hf-6-5'
e4fde8c
-
Add Inside Out emotion chat app built with Gradio
ClaudeA warm, chat-based Gradio app where multiple emotion agents (Joy, Sadness, Fear, Anger, Disgust, Anxiety, Envy, Ennui, Embarrassment) chime in on the user's message to help them discover their feelings. - Orchestrator agent picks the 2-4 most relevant emotions per message - Each emotion responds in character, in parallel, in its own voice/color - A gentle reflection agent helps the user name what is really going on - Graceful offline demo mode when no ANTHROPIC_API_KEY is set - Creative, soft, glowing 'memory orb' UI with per-emotion bubbles62d65d8
-
Initial empty commit
Clauded6d9389 -
Merge pull request #1 from peteryej/claude/trusting-galileo-o1XOS
Unverified signatureInside Out — multi-agent emotion chat app (Gradio)d782065 -
initial commit
Verified signature853182d -
Duplicate from gradio-templates/chatbot
Co-authored-by: Pete <pngwn@users.noreply.huggingface.co>880061d
-
Add Inside Out emotion chat app built with Gradio
ClaudeUnverified signatureA warm, chat-based Gradio app where multiple emotion agents (Joy, Sadness, Fear, Anger, Disgust, Anxiety, Envy, Ennui, Embarrassment) chime in on the user's message to help them discover their feelings. - Orchestrator agent picks the 2-4 most relevant emotions per message - Each emotion responds in character, in parallel, in its own voice/color - A gentle reflection agent helps the user name what is really going on - Graceful offline demo mode when no ANTHROPIC_API_KEY is set - Creative, soft, glowing 'memory orb' UI with per-emotion bubbles6249e2a
-
Initial empty commit
ClaudeUnverified signaturec1f4e7a