Views
No views yet
llama.cpp, Ollama, LM Studio, Jan, dan runtime GGUF lainnya.| File | Kuantisasi | Ukuran | Rekomendasi |
|---|---|---|---|
gemma-3-1b-it.F16.gguf | F16 | 1.87 GB | 🔬 Reference (full precision) |
gemma-3-1b-it.Q4_K_M.gguf | Q4_K_M | 0.75 GB | 💾 RAM terbatas (8GB) |
gemma-3-1b-it.Q5_K_M.gguf | Q5_K_M | 0.79 GB | ⭐ Recommended (balance) |
gemma-3-1b-it.Q8_0.gguf | Q8_0 | 1.00 GB | 🎯 Hampir lossless |
get_stock(sku, warehouse_id) - Cek stok SKU di warehousetransfer_stock(sku, quantity, source_warehouse_id, destination_warehouse_id) - Transfer stoklist_warehouses(region=None) - List semua warehouseget_stock_history(sku, warehouse_id, days=30) - Riwayat stokcreate_purchase_order(sku, quantity, supplier_id, warehouse_id) - Buat POget_supplier_info(supplier_id) - Info suppliercheck_low_stock_alerts(warehouse_id) - Alert stok rendahget_warehouse_capacity(warehouse_id) - Kapasitas warehousesearch_sku_by_name(query) - Cari SKU by namaget_shipping_status(order_id) - Status pengirimanestimate_delivery_time(source_warehouse_id, destination) - ETA pengirimanget_supplier_lead_time(supplier_id) - Lead time supplier1<think>
2[Reasoning block - optional untuk query kompleks]
3</think>
4<tool_call>
5{"name": "get_stock", "arguments": {"sku": "ZX-1042", "warehouse_id": "JKT-01"}}
6</tool_call>1./llama-cli -m warehouseflow-gemma3-1b-gguf/gemma-3-1b-it.Q5_K_M.gguf \
2 -p "Cek stok ZX-1042 di JKT-01" \
3 -n 256 --temp 0Modelfile:1FROM hf.co/Ugisr/warehouseflow-gemma3-1b-it-gguf:F16
2PARAMETER temperature 0 PARAMETER num_ctx 2048 PARAMETER stop "<end_of_turn>"
3TEMPLATE """<start_of_turn>user {{ .System }}
4{{ .Prompt }}<end_of_turn> <start_of_turn>model """
5SYSTEM """You are a warehouse management AI assistant with access to these tools:
6 1. get_stock(sku: str, warehouse_id: str)
7 2. transfer_stock(sku: str, quantity: int, source_warehouse_id: str, destination_warehouse_id: str)
8 3. list_warehouses(region: str = None)
9 4. get_stock_history(sku: str, warehouse_id: str, days: int = 30)
10 5. create_purchase_order(sku: str, quantity: int, supplier_id: str, warehouse_id: str)
11 6. get_supplier_info(supplier_id: str)
12 7. check_low_stock_alerts(warehouse_id: str)
13 8. get_warehouse_capacity(warehouse_id: str)
14 9. search_sku_by_name(query: str)
15 10. get_shipping_status(order_id: str)
16 11. estimate_delivery_time(source_warehouse_id: str, destination: str)
17 12. get_supplier_lead_time(supplier_id: str)
18Reasoning & output rules:
19 • For a simple single lookup, answer with exactly one tool call and no reasoning block.
20 • For planning, several independent lookups, or conditional decisions, FIRST write reasoning inside ..., then act.
21 • You may emit one or more <tool_call>...</tool_call> blocks. Emit several ONLY when calls are independent and every argument is already known.
22 • Each <tool_call> must contain exactly one JSON object: {"name": "", "arguments": }.
23 • If a decision depends on a value you do not have yet, call the tool that fetches it FIRST; never guess.
24 • If the request is ambiguous, lacks required info, or is outside warehouse operations, do NOT emit any tool call; ask a short clarification in plain text.
25 • Only use the tools listed above. Never invent a tool name or an argument value.
26 • Respond in the same language as the user's query."""1ollama create warehouseflow -f Modelfile
2ollama run warehouseflow "Cek stok ZX-1042 di JKT-01."1from llama_cpp import Llama
2
3llm = Llama(
4 model_path="gemma-3-1b-it.Q5_K_M.gguf",
5 n_ctx=2048,
6 n_gpu_layers=-1, # gunakan GPU jika ada
7)
8
9output = llm(
10 "Cek stok ZX-1042 di JKT-01.",
11 max_tokens=256,
12 temperature=0,
13)
14print(output["choices"][0]["text"]).gguf ke folder models| Metrik | Hasil |
|---|---|
| AST & Schema Match | 89.6% |
| OOD Refusal | 95.0% |
| Error Recovery | 100.0% |
| Trajectory Success | 100.0% |
<think> sebelum memanggil tool