From f6ef3cb2237d8c336e915ef77ddbe6f37934c4fd Mon Sep 17 00:00:00 2001
From: MIDORIBIN <aquarion123@gmail.com>
Date: Fri, 28 Jul 2023 19:07:17 +0900
Subject: refactor: refactor provider

---
 tool/provider_init.py |  33 ++++++++++++++++
 tool/readme_table.py  | 103 ++++++++++++++++++++++++++++++++++++++++++++++++++
 tool/vercel.py        | 103 ++++++++++++++++++++++++++++++++++++++++++++++++++
 3 files changed, 239 insertions(+)
 create mode 100644 tool/provider_init.py
 create mode 100644 tool/readme_table.py
 create mode 100644 tool/vercel.py

(limited to 'tool')

diff --git a/tool/provider_init.py b/tool/provider_init.py
new file mode 100644
index 00000000..22f21d4d
--- /dev/null
+++ b/tool/provider_init.py
@@ -0,0 +1,33 @@
+from pathlib import Path
+
+
+def main():
+    content = create_content()
+    with open("g4f/provider/__init__.py", "w", encoding="utf-8") as f:
+        f.write(content)
+
+
+def create_content():
+    path = Path()
+    paths = path.glob("g4f/provider/*.py")
+    paths = [p for p in paths if p.name not in ["__init__.py", "base_provider.py"]]
+    classnames = [p.stem for p in paths]
+
+    import_lines = [f"from .{name} import {name}" for name in classnames]
+    import_content = "\n".join(import_lines)
+
+    classnames.insert(0, "BaseProvider")
+    all_content = [f'    "{name}"' for name in classnames]
+    all_content = ",\n".join(all_content)
+    all_content = f"__all__ = [\n{all_content},\n]"
+
+    return f"""from .base_provider import BaseProvider
+{import_content}
+
+
+{all_content}
+"""
+
+
+if __name__ == "__main__":
+    main()
diff --git a/tool/readme_table.py b/tool/readme_table.py
new file mode 100644
index 00000000..a4d07adf
--- /dev/null
+++ b/tool/readme_table.py
@@ -0,0 +1,103 @@
+import re
+import sys
+from pathlib import Path
+from urllib.parse import urlparse
+
+sys.path.append(str(Path(__file__).parent.parent))
+
+from g4f import models, provider
+from g4f.provider.base_provider import BaseProvider
+
+
+def main():
+    print_providers()
+    print("\n", "-" * 50, "\n")
+    print_models()
+
+
+def print_providers():
+    lines = [
+        "| Website| Provider| gpt-3.5 | gpt-4 | Streaming | Status | Auth |",
+        "| ------ | ------- | ------- | ----- | --------- | ------ | ---- |",
+    ]
+    providers = get_providers()
+    for _provider in providers:
+        netloc = urlparse(_provider.url).netloc
+        website = f"[{netloc}]({_provider.url})"
+
+        provider_name = f"g4f.provider.{_provider.__name__}"
+
+        has_gpt_35 = "✔️" if _provider.supports_gpt_35_turbo else "❌"
+        has_gpt_4 = "✔️" if _provider.supports_gpt_4 else "❌"
+        stream = "✔️" if _provider.supports_stream else "❌"
+        status = (
+            "![Active](https://img.shields.io/badge/Active-brightgreen)"
+            if _provider.working
+            else "![Inactive](https://img.shields.io/badge/Inactive-red)"
+        )
+        auth = "✔️" if _provider.needs_auth else "❌"
+
+        lines.append(
+            f"| {website} | {provider_name} | {has_gpt_35} | {has_gpt_4} | {stream} | {status} | {auth} |"
+        )
+    print("\n".join(lines))
+
+
+def get_providers() -> list[type[BaseProvider]]:
+    provider_names = dir(provider)
+    ignore_names = [
+        "base_provider",
+        "BaseProvider",
+    ]
+    provider_names = [
+        provider_name
+        for provider_name in provider_names
+        if not provider_name.startswith("__") and provider_name not in ignore_names
+    ]
+    return [getattr(provider, provider_name) for provider_name in provider_names]
+
+
+def print_models():
+    base_provider_names = {
+        "cohere": "Cohere",
+        "google": "Google",
+        "openai": "OpenAI",
+        "anthropic": "Anthropic",
+        "replicate": "Replicate",
+        "huggingface": "Huggingface",
+    }
+    provider_urls = {
+        "Bard": "https://bard.google.com/",
+        "H2o": "https://www.h2o.ai/",
+        "Vercel": "https://sdk.vercel.ai/",
+    }
+
+    lines = [
+        "| Model | Base Provider | Provider | Website |",
+        "| ----- | ------------- | -------- | ------- |",
+    ]
+
+    _models = get_models()
+    for model in _models:
+        split_name = re.split(r":|/", model.name)
+        name = split_name[-1]
+
+        base_provider = base_provider_names[model.base_provider]
+        provider_name = f"g4f.provider.{model.best_provider.__name__}"
+
+        provider_url = provider_urls[model.best_provider.__name__]
+        netloc = urlparse(provider_url).netloc
+        website = f"[{netloc}]({provider_url})"
+        lines.append(f"| {name} | {base_provider} | {provider_name} | {website} |")
+
+    print("\n".join(lines))
+
+
+def get_models():
+    _models = [item[1] for item in models.__dict__.items()]
+    _models = [model for model in _models if type(model) is models.Model]
+    return [model for model in _models if model.name not in ["gpt-3.5-turbo", "gpt-4"]]
+
+
+if __name__ == "__main__":
+    main()
diff --git a/tool/vercel.py b/tool/vercel.py
new file mode 100644
index 00000000..7b87e298
--- /dev/null
+++ b/tool/vercel.py
@@ -0,0 +1,103 @@
+import json
+import re
+from typing import Any
+
+import quickjs
+from curl_cffi import requests
+
+session = requests.Session(impersonate="chrome107")
+
+
+def get_model_info() -> dict[str, Any]:
+    url = "https://sdk.vercel.ai"
+    response = session.get(url)
+    html = response.text
+    paths_regex = r"static\/chunks.+?\.js"
+    separator_regex = r'"\]\)<\/script><script>self\.__next_f\.push\(\[.,"'
+
+    paths = re.findall(paths_regex, html)
+    paths = [re.sub(separator_regex, "", path) for path in paths]
+    paths = list(set(paths))
+
+    urls = [f"{url}/_next/{path}" for path in paths]
+    scripts = [session.get(url).text for url in urls]
+
+    for script in scripts:
+        models_regex = r'let .="\\n\\nHuman:\",r=(.+?),.='
+        matches = re.findall(models_regex, script)
+
+        if matches:
+            models_str = matches[0]
+            stop_sequences_regex = r"(?<=stopSequences:{value:\[)\D(?<!\])"
+            models_str = re.sub(
+                stop_sequences_regex, re.escape('"\\n\\nHuman:"'), models_str
+            )
+
+            context = quickjs.Context()  # type: ignore
+            json_str: str = context.eval(f"({models_str})").json()  # type: ignore
+            return json.loads(json_str)  # type: ignore
+
+    return {}
+
+
+def convert_model_info(models: dict[str, Any]) -> dict[str, Any]:
+    model_info: dict[str, Any] = {}
+    for model_name, params in models.items():
+        default_params = params_to_default_params(params["parameters"])
+        model_info[model_name] = {"id": params["id"], "default_params": default_params}
+    return model_info
+
+
+def params_to_default_params(parameters: dict[str, Any]):
+    defaults: dict[str, Any] = {}
+    for key, parameter in parameters.items():
+        if key == "maximumLength":
+            key = "maxTokens"
+        defaults[key] = parameter["value"]
+    return defaults
+
+
+def get_model_names(model_info: dict[str, Any]):
+    model_names = model_info.keys()
+    model_names = [
+        name
+        for name in model_names
+        if name not in ["openai:gpt-4", "openai:gpt-3.5-turbo"]
+    ]
+    model_names.sort()
+    return model_names
+
+
+def print_providers(model_names: list[str]):
+    for name in model_names:
+        split_name = re.split(r":|/", name)
+        base_provider = split_name[0]
+        variable_name = split_name[-1].replace("-", "_").replace(".", "")
+        line = f'{variable_name} = Model(name="{name}", base_provider="{base_provider}", best_provider=Vercel,)\n'
+        print(line)
+
+
+def print_convert(model_names: list[str]):
+    for name in model_names:
+        split_name = re.split(r":|/", name)
+        key = split_name[-1]
+        variable_name = split_name[-1].replace("-", "_").replace(".", "")
+        # "claude-instant-v1": claude_instant_v1,
+        line = f'        "{key}": {variable_name},'
+        print(line)
+
+
+def main():
+    model_info = get_model_info()
+    model_info = convert_model_info(model_info)
+    print(json.dumps(model_info, indent=2))
+
+    model_names = get_model_names(model_info)
+    print("-------" * 40)
+    print_providers(model_names)
+    print("-------" * 40)
+    print_convert(model_names)
+
+
+if __name__ == "__main__":
+    main()
-- 
cgit v1.2.3