add version flag to CLI.

.gitignore

@@ -12,3 +12,4 @@ dist/
 llmsql_workdir
 
 evaluation_*
+coverage.xml

examples/inference_transformers.ipynb

@@ -87,7 +87,7 @@
  ],
  "metadata": {
   "kernelspec": {
-   "display_name": "llmsql-benchmark-3.11",
+   "display_name": "llmsql-benchmark-3.11 (3.11.13)",
    "language": "python",
    "name": "python3"
   },

examples/inference_transformers_version_1.0.ipynb

@@ -0,0 +1,85 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": 1,
+   "id": "e71979fc",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "/home/pihul/Desktop/gdrive/Projects/llmsql-benchmark/.venv/lib/python3.11/site-packages/tqdm/auto.py:21: TqdmWarning: IProgress not found. Please update jupyter and ipywidgets. See https://ipywidgets.readthedocs.io/en/stable/user_install.html\n",
+      "  from .autonotebook import tqdm as notebook_tqdm\n"
+     ]
+    },
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "Loading model from: EleutherAI/pythia-14m\n"
+     ]
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "`torch_dtype` is deprecated! Use `dtype` instead!\n",
+      "2026-02-23 09:00:00,424 [INFO] llmsql-bench: Removing existing path: llmsql_workdir/questions.jsonl\n",
+      "2026-02-23 09:00:00,426 [INFO] llmsql-bench: Downloading questions.jsonl from Hugging Face Hub...\n",
+      "2026-02-23 09:00:01,433 [INFO] llmsql-bench: Downloaded questions.jsonl to: llmsql_workdir/questions.jsonl\n",
+      "2026-02-23 09:00:01,434 [INFO] llmsql-bench: Removing existing path: llmsql_workdir/tables.jsonl\n",
+      "2026-02-23 09:00:01,436 [INFO] llmsql-bench: Downloading tables.jsonl from Hugging Face Hub...\n",
+      "2026-02-23 09:00:02,117 [INFO] llmsql-bench: Downloaded tables.jsonl to: llmsql_workdir/tables.jsonl\n",
+      "2026-02-23 09:00:03,373 [INFO] llmsql-bench: Limiting evaluation to first 100 questions out of 80330\n",
+      "2026-02-23 09:00:03,383 [INFO] llmsql-bench: Writing results to test_output.jsonl\n",
+      "2026-02-23 09:00:03,384 [INFO] llmsql-bench: Using 5-shot prompt builder: build_prompt_5shot\n",
+      "Generating:   0%|          | 0/4 [00:00<?, ?it/s]Asking to truncate to max_length but no maximum length is provided and the model has no predefined maximum length. Default to no truncation.\n",
+      "The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.\n",
+      "2026-02-23 09:00:10,474 [INFO] llmsql-bench: Saved batch 1: 25/100\n",
+      "Generating:  25%|██▌       | 1/4 [00:07<00:21,  7.09s/it]2026-02-23 09:00:19,324 [INFO] llmsql-bench: Saved batch 2: 50/100\n",
+      "Generating:  50%|█████     | 2/4 [00:15<00:16,  8.12s/it]2026-02-23 09:00:28,513 [INFO] llmsql-bench: Saved batch 3: 75/100\n",
+      "Generating:  75%|███████▌  | 3/4 [00:25<00:08,  8.61s/it]2026-02-23 09:00:37,861 [INFO] llmsql-bench: Saved batch 4: 100/100\n",
+      "Generating: 100%|██████████| 4/4 [00:34<00:00,  8.62s/it]\n",
+      "2026-02-23 09:00:37,864 [INFO] llmsql-bench: Generation complete — total: 100\n"
+     ]
+    }
+   ],
+   "source": [
+    "from llmsql import inference_transformers\n",
+    "\n",
+    "# Example 1: Basic usage (same as before)\n",
+    "results = inference_transformers(\n",
+    "    model_or_model_name_or_path=\"EleutherAI/pythia-14m\",\n",
+    "    output_file=\"test_output.jsonl\",\n",
+    "    batch_size=25,\n",
+    "    do_sample=False,\n",
+    "    limit=100,\n",
+    "    version=\"1.0\"\n",
+    "    )\n"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "llmsql-benchmark-3.11 (3.11.13)",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.11.13"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}

llmsql/__main__.py

@@ -1,155 +1,11 @@
-import argparse
-import inspect
-import json
+from llmsql._cli import ParserCLI
 
 
 def main() -> None:
-    parser = argparse.ArgumentParser(prog="llmsql", description="LLMSQL CLI")
-    subparsers = parser.add_subparsers(dest="command")
-
-    # ================================================================
-    # Inference command
-    # ================================================================
-    inference_examples = r"""
-Examples:
-
-  # 1️⃣ Run inference with Transformers backend
-  llmsql inference --method transformers \
-      --model-or-model-name-or-path Qwen/Qwen2.5-1.5B-Instruct \
-      --output-file outputs/preds_transformers.jsonl \
-      --batch-size 8 \
-      --num-fewshots 5
-
-  # 2️⃣ Run inference with vLLM backend
-  llmsql inference --method vllm \
-      --model-name Qwen/Qwen2.5-1.5B-Instruct \
-      --output-file outputs/preds_vllm.jsonl \
-      --batch-size 8 \
-      --num-fewshots 5
-
-  # 3️⃣ Pass model-specific kwargs (for Transformers)
-  llmsql inference --method transformers \
-      --model-or-model-name-or-path meta-llama/Llama-3-8b-instruct \
-      --output-file outputs/llama_preds.jsonl \
-      --model-kwargs '{"attn_implementation": "flash_attention_2", "torch_dtype": "bfloat16"}'
-
-  # 4️⃣ Pass LLM init kwargs (for vLLM)
-  llmsql inference --method vllm \
-      --model-name mistralai/Mixtral-8x7B-Instruct-v0.1 \
-      --output-file outputs/mixtral_preds.jsonl \
-      --llm-kwargs '{"max_model_len": 4096, "gpu_memory_utilization": 0.9}'
-
-  # 5️⃣ Override generation parameters dynamically
-  llmsql inference --method transformers \
-      --model-or-model-name-or-path Qwen/Qwen2.5-1.5B-Instruct \
-      --output-file outputs/temp_0.9.jsonl \
-      --temperature 0.9 \
-      --generation-kwargs '{"do_sample": true, "top_p": 0.9, "top_k": 40}'
-"""
-
-    inf_parser = subparsers.add_parser(
-        "inference",
-        help="Run inference using either Transformers or vLLM backend.",
-        description="Run SQL generation using a chosen inference method "
-        "(either 'transformers' or 'vllm').",
-        epilog=inference_examples,
-        formatter_class=argparse.RawTextHelpFormatter,
-    )
-
-    inf_parser.add_argument(
-        "--method",
-        type=str,
-        required=True,
-        choices=["transformers", "vllm"],
-        help="Inference backend to use ('transformers' or 'vllm').",
-    )
-
-    # ================================================================
-    # Parse CLI
-    # ================================================================
-    args, extra = parser.parse_known_args()
-
-    # ------------------------------------------------
-    # Inference
-    # ------------------------------------------------
-    if args.command == "inference":
-        if args.method == "vllm":
-            from llmsql import inference_vllm as inference_fn
-        elif args.method == "transformers":
-            from llmsql import inference_transformers as inference_fn  # type: ignore
-        else:
-            raise ValueError(f"Unknown inference method: {args.method}")
-
-        # Dynamically create parser from the function signature
-        fn_parser = argparse.ArgumentParser(
-            prog=f"llmsql inference --method {args.method}",
-            description=f"Run inference using {args.method} backend",
-        )
-
-        sig = inspect.signature(inference_fn)
-        for name, param in sig.parameters.items():
-            if param.kind == inspect.Parameter.VAR_KEYWORD:
-                fn_parser.add_argument(
-                    "--llm-kwargs",
-                    default="{}",
-                    help="Additional LLM kwargs as a JSON string, e.g. '{\"top_p\": 0.9}'",
-                )
-                fn_parser.add_argument(
-                    "--generate-kwargs",
-                    default="{}",
-                    help="",
-                )
-                continue
-            arg_name = f"--{name.replace('_', '-')}"
-            default = param.default
-            if default is inspect.Parameter.empty:
-                fn_parser.add_argument(arg_name, required=True)
-            else:
-                if isinstance(default, bool):
-                    fn_parser.add_argument(
-                        arg_name,
-                        action="store_true" if not default else "store_false",
-                        help=f"(default: {default})",
-                    )
-                elif default is None:
-                    fn_parser.add_argument(arg_name, type=str, default=None)
-                else:
-                    fn_parser.add_argument(
-                        arg_name, type=type(default), default=default
-                    )
-
-        fn_args = fn_parser.parse_args(extra)
-        fn_kwargs = vars(fn_args)
-
-        if "llm_kwargs" in fn_kwargs and isinstance(fn_kwargs["llm_kwargs"], str):
-            try:
-                fn_kwargs["llm_kwargs"] = json.loads(fn_kwargs["llm_kwargs"])
-            except json.JSONDecodeError:
-                print("⚠️  Could not parse --llm-kwargs JSON, passing as string.")
-
-        if fn_kwargs.get("model_kwargs") is not None:
-            try:
-                fn_kwargs["model_kwargs"] = json.loads(fn_kwargs["model_kwargs"])
-            except json.JSONDecodeError:
-                raise
-
-        if fn_kwargs.get("generation_kwargs") is not None:
-            try:
-                fn_kwargs["generation_kwargs"] = json.loads(
-                    fn_kwargs["generation_kwargs"]
-                )
-            except json.JSONDecodeError:
-                raise
-
-        print(f"🔹 Running {args.method} inference with arguments:")
-        for k, v in fn_kwargs.items():
-            print(f"  {k}: {v}")
-
-        results = inference_fn(**fn_kwargs)
-        print(f"✅ Inference complete. Generated {len(results)} results.")
-
-    else:
-        parser.print_help()
+    """Main CLI entry point."""
+    parser = ParserCLI()
+    args = parser.parse_args()
+    parser.execute(args)
 
 
 if __name__ == "__main__":

llmsql/_cli/__init__.py

@@ -0,0 +1,7 @@
+"""
+CLI subcommands to run from the terminal.
+"""
+
+from .llmsql_cli import ParserCLI
+
+__all__ = ["ParserCLI"]