fix a bunch of image uri stuff

Author: iamlemec

oneping/curl.py

@@ -71,6 +71,7 @@ def prepare_request(
     payload = {**payload_model, **payload_message, **kwargs}
 
     # add in max tokens
+    max_tokens = max_tokens if max_tokens is not None else prov.get('max_tokens_default')
     if max_tokens is not None:
         payload[prov.max_tokens_name] = max_tokens
 

oneping/native/google.py

@@ -6,7 +6,7 @@
 from google.genai.types import Part, Content, GenerateContentConfig
 
 from ..providers import CONFIG as C, PROVIDERS as P
-from ..utils import parse_image_uri
+from ..utils import ensure_image_uri, parse_image_uri
 
 ##
 ## helper functions
@@ -15,7 +15,8 @@
 def make_content(text, image=None):
     parts = [Part(text=text)]
     if image is not None:
-        mime_type, data = parse_image_uri(image)
+        image_url = ensure_image_uri(image)
+        mime_type, data = parse_image_uri(image_url)
         part = Part.from_bytes(data=data, mime_type=mime_type)
         parts.append(part)
     return parts

oneping/native/openai.py

@@ -7,7 +7,7 @@
     CONFIG as C, PROVIDERS as P,
     content_openai, convert_history, payload_openai,
     response_openai_native, stream_openai_native,
-    embed_response_openai, transcribe_response_openai
+    embed_response_openai_native, transcribe_response_openai
 )
 
 ##
@@ -56,8 +56,8 @@ async def stream_async(query, image=None, history=None, prefill=None, prediction
 
 def embed(query, model=P.openai.embed_model, api_key=None, base_url=None, **kwargs):
     client = make_client(base_url=base_url, api_key=api_key)
-    response = client.embeddings.create(query, model=model, **kwargs)
-    return embed_response_openai(response)
+    response = client.embeddings.create(input=query, model=model, **kwargs)
+    return embed_response_openai_native(response)
 
 def transcribe(audio, model=P.openai.transcribe_model, api_key=None, base_url=None, **kwargs):
     client = make_client(base_url=base_url, api_key=api_key)

oneping/providers.py

@@ -4,7 +4,7 @@
 import toml
 from pathlib import Path
 
-from .utils import split_image_uri, Config
+from .utils import split_image_uri, ensure_image_uri, Config
 
 ##
 ## authorization headers
@@ -17,7 +17,7 @@ def authorize_openai(api_key):
 
 def authorize_anthropic(api_key):
     return {
-        'x-api-key': api_key,
+        'X-Api-Key': api_key,
     }
 
 ##
@@ -27,7 +27,7 @@ def authorize_anthropic(api_key):
 def content_openai(text, image=None):
     if image is None:
         return text
-    image_url = { 'url': image }
+    image_url = { 'url': ensure_image_uri(image) }
     return [
         { 'type': 'image_url', 'image_url': image_url },
         { 'type': 'text', 'text': text },
@@ -36,7 +36,8 @@ def content_openai(text, image=None):
 def content_anthropic(text, image=None):
     if image is None:
         return text
-    media_type, data = split_image_uri(image)
+    image_url = ensure_image_uri(image)
+    media_type, data = split_image_uri(image_url)
     source = {
         'type': 'base64', 'media_type': media_type, 'data': data
     }
@@ -174,6 +175,11 @@ def embed_response_openai(reply):
         item['embedding'] for item in reply['data']
     ]
 
+def embed_response_openai_native(reply):
+    return [
+        item.embedding for item in reply.data
+    ]
+
 def embed_payload_tei(text):
     return {'inputs': text}
 
@@ -264,22 +270,18 @@ def transcribe_response_openai(audio):
 ## known llm providers
 ##
 
-# get config paths
-library_dir = Path(__file__).parent
-xdg_config_home = Path(os.environ.get('XDG_CONFIG_HOME', os.path.expanduser('~/.config')))
-user_config_dir = xdg_config_home / 'oneping'
-default_providers_file = library_dir / 'providers.toml'
-user_providers_file = user_config_dir / 'providers.toml'
-default_config_file = library_dir / 'config.toml'
-user_config_file = user_config_dir / 'config.toml'
-
 # fault tolerant toml loader
 def load_toml(file):
     if os.path.exists(file):
         return toml.load(file)
     else:
         return {}
 
+# get config paths
+XDG_LOC = os.path.expanduser('~/.config')
+LIB_DIR = Path(__file__).parent
+XDG_DIR = Path(os.environ.get('XDG_CONFIG_HOME', XDG_LOC))
+
 # merge config layers
 global PROVIDERS
 global CONFIG
@@ -288,10 +290,10 @@ def reload():
     global CONFIG
 
     # reload config from disk
-    DEFAULT_CONFIG = load_toml(default_config_file)
-    USER_CONFIG = load_toml(user_config_file)
-    DEFAULT_PROVIDERS = load_toml(default_providers_file)
-    USER_PROVIDERS = load_toml(user_providers_file)
+    DEFAULT_CONFIG = load_toml(LIB_DIR / 'config.toml')
+    DEFAULT_PROVIDERS = load_toml(LIB_DIR / 'providers.toml')
+    USER_CONFIG = load_toml(XDG_DIR / 'oneping' / 'config.toml')
+    USER_PROVIDERS = load_toml(XDG_DIR / 'oneping' / 'providers.toml')
 
     # merge provider layers
     CONFIG = Config({ **DEFAULT_CONFIG, **USER_CONFIG })

oneping/providers.toml

@@ -46,44 +46,53 @@ embed_model = "text-embedding-3-large"
 [anthropic]
 base_url = "https://api.anthropic.com"
 authorize = "anthropic"
-chat_path = "messages"
+chat_path = "v1/messages"
 max_tokens_name = "max_tokens"
+max_tokens_default = 8192
 content = "anthropic"
 payload = "anthropic"
 response = "anthropic"
 stream = "anthropic"
 api_key_env = "ANTHROPIC_API_KEY"
-chat_model = "claude-opus-4-1-20250805"
+chat_model = "claude-sonnet-4-5-20250929"
 
 [anthropic.headers]
 anthropic-version = "2023-06-01"
 
 [google]
-base_url = "https://generativelanguage.googleapis.com/v1beta/openai"
+base_url = "https://generativelanguage.googleapis.com/v1beta"
+authorize = "openai"
+chat_path = "openai/chat/completions"
+embed_path = "openai/embeddings"
 api_key_env = "GEMINI_API_KEY"
-chat_model = "gemini-2.0-flash-exp"
-embed_model = "gemini-embedding-exp-03-07"
+chat_model = "gemini-2.5-flash"
+embed_model = "gemini-embedding-001"
 
 [xai]
 base_url = "https://api.x.ai/v1"
+authorize = "openai"
 api_key_env = "XAI_API_KEY"
 chat_model = "grok-4"
 
 [fireworks]
 base_url = "https://api.fireworks.ai/inference"
+authorize = "openai"
 api_key_env = "FIREWORKS_API_KEY"
 chat_model = "accounts/fireworks/models/llama-v3p3-70b-instruct"
 
 [groq]
 base_url = "https://api.groq.com/openai"
+authorize = "openai"
 api_key_env = "GROQ_API_KEY"
 chat_model = "llama-3.3-70b-versatile"
 
 [deepseek]
 base_url = "https://api.deepseek.com"
+authorize = "openai"
 api_key_env = "DEEPSEEK_API_KEY"
 chat_model = "deepseek-chat"
 
 [azure]
 api_key_env = "AZURE_API_KEY"
+authorize = "openai"
 chat_model = "gpt-5"