Spaces:
Paused
Paused
Upload patch_models.py
Browse files- patch_models.py +7 -8
patch_models.py
CHANGED
|
@@ -8,13 +8,13 @@ with open(IDS_FILE, "r") as f:
|
|
| 8 |
# Add new model ID constants after existing ones
|
| 9 |
old_models_block = 'CLAUDE_OPUS_48_MODEL_ID = "anthropic/claude-opus-4.8:fal-ai"\nGPT_55_MODEL_ID = "openai/gpt-5.5:fal-ai"\nKIMI_K27_CODE_MODEL_ID = "moonshotai/Kimi-K2.7-Code:novita"\nMINIMAX_M3_MODEL_ID = "MiniMaxAI/MiniMax-M3:novita"\nGLM_52_MODEL_ID = "zai-org/GLM-5.2:novita"\nDEEPSEEK_V4_PRO_MODEL_ID = "deepseek-ai/DeepSeek-V4-Pro:novita"'
|
| 10 |
|
| 11 |
-
new_models_block = 'GPT_55_MODEL_ID = "openai/gpt-5.5:fal-ai"\nKIMI_K27_CODE_MODEL_ID = "deepseek-ai/DeepSeek-V4-Pro"\nMINIMAX_M3_MODEL_ID = "deepseek-ai/DeepSeek-V4-Flash"\nGLM_52_MODEL_ID = "openai/deepseek/deepseek-v4-flash"\nDEEPSEEK_V4_PRO_MODEL_ID = "nvidia/nemotron-3-super-120b-a12b:free"\n\n# === PATCH: New model IDs ===\
|
| 12 |
|
| 13 |
content = content.replace(old_models_block, new_models_block)
|
| 14 |
|
| 15 |
# Also update the HOSTED_MODEL_IDS set
|
| 16 |
old_hosted = "HOSTED_MODEL_IDS = {\n CLAUDE_OPUS_48_MODEL_ID,\n GPT_55_MODEL_ID,\n KIMI_K27_CODE_MODEL_ID,\n MINIMAX_M3_MODEL_ID,\n GLM_52_MODEL_ID,\n DEEPSEEK_V4_PRO_MODEL_ID,\n}"
|
| 17 |
-
new_hosted = "HOSTED_MODEL_IDS = {\n GPT_55_MODEL_ID,\n KIMI_K27_CODE_MODEL_ID,\n MINIMAX_M3_MODEL_ID,\n GLM_52_MODEL_ID,\n DEEPSEEK_V4_PRO_MODEL_ID,\n
|
| 18 |
content = content.replace(old_hosted, new_hosted)
|
| 19 |
|
| 20 |
with open(IDS_FILE, "w") as f:
|
|
@@ -26,18 +26,17 @@ AGENT_FILE = "/app/backend/routes/agent.py"
|
|
| 26 |
with open(AGENT_FILE, "r") as f:
|
| 27 |
content = f.read()
|
| 28 |
|
| 29 |
-
# Update imports
|
| 30 |
old_import = "from agent.core.model_ids import (\n CLAUDE_OPUS_48_MODEL_ID,\n DEEPSEEK_V4_PRO_MODEL_ID,\n GLM_52_MODEL_ID,\n GPT_55_MODEL_ID,\n KIMI_K27_CODE_MODEL_ID,\n MINIMAX_M3_MODEL_ID,\n strip_huggingface_model_prefix,\n)"
|
| 31 |
-
new_import = "from agent.core.model_ids import (\n DEEPSEEK_V4_PRO_MODEL_ID,\n GLM_52_MODEL_ID,\n GPT_55_MODEL_ID,\n KIMI_K27_CODE_MODEL_ID,\n MINIMAX_M3_MODEL_ID,\n
|
| 32 |
content = content.replace(old_import, new_import)
|
| 33 |
|
| 34 |
-
# Expand _available_models() —
|
| 35 |
old_available = 'def _available_models() -> list[dict[str, Any]]:\n models = [\n {\n "id": CLAUDE_OPUS_48_MODEL_ID,\n "label": "Claude Opus 4.8",\n },\n {\n "id": DEFAULT_GPT_MODEL_ID,\n "label": "GPT-5.5",\n },\n {\n "id": KIMI_K27_CODE_MODEL_ID,\n "label": "Kimi K2.7 Code",\n },\n {\n "id": MINIMAX_M3_MODEL_ID,\n "label": "MiniMax M3",\n },\n {\n "id": DEFAULT_MODEL_ID,\n "label": "GLM 5.2",\n "recommended": True,\n },\n {\n "id": DEEPSEEK_V4_PRO_MODEL_ID,\n "label": "DeepSeek V4 Pro",\n },\n ]\n return models'
|
| 36 |
-
new_available = 'def _available_models() -> list[dict[str, Any]]:\n models = [\n {\n "id":
|
| 37 |
content = content.replace(old_available, new_available)
|
| 38 |
|
| 39 |
-
# Update title generation
|
| 40 |
-
# Change it to use our HF router model
|
| 41 |
old_title = '"openai/gpt-oss-120b:cerebras",'
|
| 42 |
new_title = '"huggingface/deepseek-ai/DeepSeek-V4-Pro",'
|
| 43 |
content = content.replace(old_title, new_title)
|
|
|
|
| 8 |
# Add new model ID constants after existing ones
|
| 9 |
old_models_block = 'CLAUDE_OPUS_48_MODEL_ID = "anthropic/claude-opus-4.8:fal-ai"\nGPT_55_MODEL_ID = "openai/gpt-5.5:fal-ai"\nKIMI_K27_CODE_MODEL_ID = "moonshotai/Kimi-K2.7-Code:novita"\nMINIMAX_M3_MODEL_ID = "MiniMaxAI/MiniMax-M3:novita"\nGLM_52_MODEL_ID = "zai-org/GLM-5.2:novita"\nDEEPSEEK_V4_PRO_MODEL_ID = "deepseek-ai/DeepSeek-V4-Pro:novita"'
|
| 10 |
|
| 11 |
+
new_models_block = 'GPT_55_MODEL_ID = "openai/gpt-5.5:fal-ai"\nKIMI_K27_CODE_MODEL_ID = "deepseek-ai/DeepSeek-V4-Pro"\nMINIMAX_M3_MODEL_ID = "deepseek-ai/DeepSeek-V4-Flash"\nGLM_52_MODEL_ID = "openai/deepseek/deepseek-v4-flash"\nDEEPSEEK_V4_PRO_MODEL_ID = "nvidia/nemotron-3-super-120b-a12b:free"\n\n# === PATCH: New model IDs (bỏ :free suffix — OpenRouter đã gỡ free variant) ===\nLLAMA_3_1_8B_MODEL_ID = "meta-llama/llama-3.1-8b-instruct"\nGEMMA_2_9B_MODEL_ID = "google/gemma-2-9b-it"\nNEMOTRON_70B_MODEL_ID = "nvidia/llama-3.1-nemotron-70b-instruct"\nQWEN3_CODER_NEXT_MODEL_ID = "Qwen/Qwen3-Coder-Next"\nGEMINI_FLASH_MODEL_ID = "openai/google/gemini-2.0-flash-001"'
|
| 12 |
|
| 13 |
content = content.replace(old_models_block, new_models_block)
|
| 14 |
|
| 15 |
# Also update the HOSTED_MODEL_IDS set
|
| 16 |
old_hosted = "HOSTED_MODEL_IDS = {\n CLAUDE_OPUS_48_MODEL_ID,\n GPT_55_MODEL_ID,\n KIMI_K27_CODE_MODEL_ID,\n MINIMAX_M3_MODEL_ID,\n GLM_52_MODEL_ID,\n DEEPSEEK_V4_PRO_MODEL_ID,\n}"
|
| 17 |
+
new_hosted = "HOSTED_MODEL_IDS = {\n GPT_55_MODEL_ID,\n KIMI_K27_CODE_MODEL_ID,\n MINIMAX_M3_MODEL_ID,\n GLM_52_MODEL_ID,\n DEEPSEEK_V4_PRO_MODEL_ID,\n LLAMA_3_1_8B_MODEL_ID,\n GEMMA_2_9B_MODEL_ID,\n NEMOTRON_70B_MODEL_ID,\n QWEN3_CODER_NEXT_MODEL_ID,\n GEMINI_FLASH_MODEL_ID,\n}"
|
| 18 |
content = content.replace(old_hosted, new_hosted)
|
| 19 |
|
| 20 |
with open(IDS_FILE, "w") as f:
|
|
|
|
| 26 |
with open(AGENT_FILE, "r") as f:
|
| 27 |
content = f.read()
|
| 28 |
|
| 29 |
+
# Update imports — remove old free models, add working free models
|
| 30 |
old_import = "from agent.core.model_ids import (\n CLAUDE_OPUS_48_MODEL_ID,\n DEEPSEEK_V4_PRO_MODEL_ID,\n GLM_52_MODEL_ID,\n GPT_55_MODEL_ID,\n KIMI_K27_CODE_MODEL_ID,\n MINIMAX_M3_MODEL_ID,\n strip_huggingface_model_prefix,\n)"
|
| 31 |
+
new_import = "from agent.core.model_ids import (\n DEEPSEEK_V4_PRO_MODEL_ID,\n GLM_52_MODEL_ID,\n GPT_55_MODEL_ID,\n KIMI_K27_CODE_MODEL_ID,\n MINIMAX_M3_MODEL_ID,\n LLAMA_3_1_8B_MODEL_ID,\n GEMMA_2_9B_MODEL_ID,\n NEMOTRON_70B_MODEL_ID,\n QWEN3_CODER_NEXT_MODEL_ID,\n GEMINI_FLASH_MODEL_ID,\n strip_huggingface_model_prefix,\n)"
|
| 32 |
content = content.replace(old_import, new_import)
|
| 33 |
|
| 34 |
+
# Expand _available_models() — Llama 3.1 8B as default (free, no :free suffix needed)
|
| 35 |
old_available = 'def _available_models() -> list[dict[str, Any]]:\n models = [\n {\n "id": CLAUDE_OPUS_48_MODEL_ID,\n "label": "Claude Opus 4.8",\n },\n {\n "id": DEFAULT_GPT_MODEL_ID,\n "label": "GPT-5.5",\n },\n {\n "id": KIMI_K27_CODE_MODEL_ID,\n "label": "Kimi K2.7 Code",\n },\n {\n "id": MINIMAX_M3_MODEL_ID,\n "label": "MiniMax M3",\n },\n {\n "id": DEFAULT_MODEL_ID,\n "label": "GLM 5.2",\n "recommended": True,\n },\n {\n "id": DEEPSEEK_V4_PRO_MODEL_ID,\n "label": "DeepSeek V4 Pro",\n },\n ]\n return models'
|
| 36 |
+
new_available = 'def _available_models() -> list[dict[str, Any]]:\n models = [\n {\n "id": LLAMA_3_1_8B_MODEL_ID,\n "label": "Llama 3.1 8B",\n "recommended": True,\n },\n {\n "id": NEMOTRON_70B_MODEL_ID,\n "label": "Nemotron 70B",\n },\n {\n "id": DEFAULT_GPT_MODEL_ID,\n "label": "GPT-5.5",\n },\n {\n "id": KIMI_K27_CODE_MODEL_ID,\n "label": "DeepSeek V4 Pro",\n },\n {\n "id": GEMMA_2_9B_MODEL_ID,\n "label": "Gemma 2 9B",\n },\n {\n "id": MINIMAX_M3_MODEL_ID,\n "label": "DeepSeek V4 Flash",\n },\n {\n "id": DEFAULT_MODEL_ID,\n "label": "DeepSeek V4 Flash",\n },\n {\n "id": DEEPSEEK_V4_PRO_MODEL_ID,\n "label": "Nemotron 3 Super 120B",\n },\n {\n "id": QWEN3_CODER_NEXT_MODEL_ID,\n "label": "Qwen3 Coder Next",\n },\n {\n "id": GEMINI_FLASH_MODEL_ID,\n "label": "Gemini 2.0 Flash",\n },\n ]\n return models'
|
| 37 |
content = content.replace(old_available, new_available)
|
| 38 |
|
| 39 |
+
# Update title generation
|
|
|
|
| 40 |
old_title = '"openai/gpt-oss-120b:cerebras",'
|
| 41 |
new_title = '"huggingface/deepseek-ai/DeepSeek-V4-Pro",'
|
| 42 |
content = content.replace(old_title, new_title)
|