From f1320b99d33b2381289cffcd1f62c83265536f01 Mon Sep 17 00:00:00 2001
From: Tyler <68524461+TySP-Dev@users.noreply.github.com>
Date: Fri, 15 May 2026 14:24:09 -0400
Subject: [PATCH] Making project Ollama focused
---
ai_answers.py | 138 +++-----------------------------------------------
1 file changed, 7 insertions(+), 131 deletions(-)
diff --git a/ai_answers.py b/ai_answers.py
index 7a5357d..ebd2e75 100644
--- a/ai_answers.py
+++ b/ai_answers.py
@@ -46,14 +46,7 @@ PLUGIN_NAME = "AI Answers"
DEFAULT_TABS = "general,science,it,news"
PROVIDER_PRESETS = {
- 'openai': {'url': 'https://api.openai.com/v1/chat/completions', 'model': 'gpt-4o-mini'},
- 'openrouter': {'url': 'https://openrouter.ai/api/v1/chat/completions', 'model': 'google/gemma-3-27b-it:free'},
'ollama': {'url': 'http://localhost:11434/v1/chat/completions', 'model': 'llama3.2'},
- 'localai': {'url': 'http://localhost:8080/v1/chat/completions', 'model': 'gpt-4'},
- 'lmstudio': {'url': 'http://localhost:1234/v1/chat/completions', 'model': 'local-model'},
- 'gemini': {'url': 'https://generativelanguage.googleapis.com/v1beta/models/{model}:streamGenerateContent', 'model': 'gemma-3-27b-it'},
- 'azure': {'url': None, 'model': 'azure-deployment'},
- 'huggingface': {'url': 'https://api-inference.huggingface.co/models/{model}/v1/chat/completions', 'model': 'meta-llama/Meta-Llama-3-8B-Instruct'}
}
# UI assets
@@ -793,37 +786,25 @@ class SXNGPlugin(Plugin):
raw_url = os.getenv('LLM_URL', '').strip()
if not raw_provider and raw_url:
url_lower = raw_url.lower()
- if 'openai.com' in url_lower:
- raw_provider = 'openai'
- elif 'openrouter.ai' in url_lower:
- raw_provider = 'openrouter'
- elif ':11434' in url_lower:
+ if ':11434' in url_lower:
raw_provider = 'ollama'
- elif 'generativelanguage.googleapis.com' in url_lower:
- raw_provider = 'gemini'
- elif 'openai.azure.com' in url_lower or '.azure.com' in url_lower:
- raw_provider = 'azure'
- elif 'huggingface.co' in url_lower:
- raw_provider = 'huggingface'
else:
- raw_provider = 'openai'
- logger.info(f"{PLUGIN_NAME}: Using OpenAI-compatible mode for custom URL")
+ raw_provider = 'error'
+ logger.info(f"{raw_provider}: Ollama not detected")
if not raw_provider:
self.provider = ''
self.model = ''
- self.is_gemini = False
self.api_key = ''
return
if raw_provider not in PROVIDER_PRESETS:
- logger.warning(f"{PLUGIN_NAME}: Unknown provider '{raw_provider}', falling back to 'openai'")
- self.provider = raw_provider if raw_provider in PROVIDER_PRESETS else 'openai'
- self.is_gemini = (self.provider == 'gemini')
+ logger.warning(f"{PLUGIN_NAME}: Not Ollama '{raw_provider}', please correct.")
+ self.provider = raw_provider if raw_provider in PROVIDER_PRESETS else 'Error'
preset = PROVIDER_PRESETS[self.provider]
self.api_key = os.getenv('LLM_KEY', '')
- if not self.api_key and self.provider in ('ollama', 'localai', 'lmstudio'):
+ if not self.api_key and self.provider == 'ollama':
self.api_key = 'none'
self.api_key = self.api_key.strip()
@@ -1121,112 +1102,7 @@ class SXNGPlugin(Plugin):
{numbered_instructions}
"""
-
- def call_gemini():
- base = self.endpoint_url.replace('streamGenerateContent', 'generateContent')
- url = f"{base}&key={self.api_key}" if '?' in base else f"{base}?key={self.api_key}"
- conn = None
- try:
- conn, path = _get_streaming_connection(url)
- payload = json.dumps({
- "contents": [{"parts": [{"text": prompt}]}],
- "generationConfig": {"maxOutputTokens": min(self.max_tokens * 4, 8192), "temperature": self.temperature}
- })
- conn.request("POST", path, body=payload.encode('utf-8'), headers={"Content-Type": "application/json"})
- res = conn.getresponse()
- if res.status != 200:
- body = res.read(2048).decode('utf-8', errors='replace')[:500]
- logger.error(f"{PLUGIN_NAME}: Gemini API {res.status}: {body}")
- return '', f"API error {res.status}. Check server logs."
- obj = json.loads(res.read().decode('utf-8', errors='replace'))
- if obj.get('promptFeedback', {}).get('blockReason'):
- return '', f"Gemini blocked prompt: {obj['promptFeedback']['blockReason']}"
- candidates = obj.get('candidates', [])
- if not candidates:
- return '', "No candidates in Gemini response."
- first = candidates[0]
- if first.get('finishReason') == 'SAFETY':
- return '', "Gemini stopped generation due to safety filters."
- parts = first.get('content', {}).get('parts', [])
- text = ''.join(p.get('text', '') for p in parts if isinstance(p, dict))
- return text, None
- except Exception as e:
- logger.error(f"{PLUGIN_NAME}: Gemini call error: {e}", exc_info=True)
- return '', f"Connection Error: {e}"
- finally:
- if conn: conn.close()
-
- def call_openai_compatible():
- conn = None
- try:
- conn, path = _get_streaming_connection(self.endpoint_url)
- payload_dict = {
- "model": effective_model,
- "messages": [
- {"role": "system", "content": SYSTEM},
- {"role": "user", "content": prompt},
- {"role": "assistant", "content": ""},
- ],
- "stream": False,
- "max_tokens": self.max_tokens,
- "temperature": self.temperature
- }
- payload = json.dumps(payload_dict)
- headers = {
- "Content-Type": "application/json",
- "HTTP-Referer": "https://github.com/searxng/searxng",
- "X-Title": "SearXNG"
- }
- if self.provider == 'azure':
- headers['api-key'] = self.api_key
- else:
- headers['Authorization'] = f"Bearer {self.api_key}"
- conn.request("POST", path, body=payload.encode('utf-8'), headers=headers)
- res = conn.getresponse()
- if res.status != 200:
- body = res.read(2048).decode('utf-8', errors='replace')[:500]
- logger.error(f"{PLUGIN_NAME}: {self.provider} API {res.status}: {body}")
- return '', f"API error {res.status}. Check server logs."
- obj = json.loads(res.read().decode('utf-8', errors='replace'))
- if "error" in obj:
- err = obj["error"]
- msg = err.get("message", str(err)) if isinstance(err, dict) else str(err)
- return '', f"API Error: {msg}"
- choices = obj.get("choices", [])
- if not choices:
- return '', "No choices in API response."
- message = choices[0].get("message", {})
- content = re.sub(r'.*?', '', message.get("content") or "", flags=re.DOTALL).strip()
- reasoning = message.get("reasoning") or message.get("reasoning_content") or ""
- if not content and reasoning:
- logger.warning(f"{PLUGIN_NAME}: {self.provider} returned empty content; extracting answer from reasoning field")
- header_pat = re.compile(r'^\s*\*?\*?[A-Z][^:]{0,40}:\*?\*?\s*$', re.MULTILINE)
- matches = list(header_pat.finditer(reasoning))
- if matches:
- answer = reasoning[matches[-1].end():].strip()
- else:
- paras = [p.strip() for p in re.split(r'\n{2,}', reasoning) if p.strip()]
- answer = paras[-1] if paras else reasoning.strip()
- full = answer
- else:
- full = (f"\n{reasoning}\n\n\n" if reasoning else "") + content
- full = re.sub(r'.*?', '', full, flags=re.DOTALL).strip()
- return full, None
- except Exception as e:
- logger.error(f"{PLUGIN_NAME}: {self.provider} call error: {e}", exc_info=True)
- return '', f"Connection Error: {e}"
- finally:
- if conn: conn.close()
-
- call_fn = call_gemini if self.is_gemini else call_openai_compatible
- text, error = call_fn()
-
- if self.provider == 'ollama' and getattr(self, 'ollama_unload_after', False):
- self._ollama_unload_model()
-
- return jsonify({"text": text, "error": error})
- return True
-
+
def _assemble_context(self, clean_results, infoboxes, answers, offset=0) -> tuple[str, list]:
"""Builds context string from normalized search data. Returns (context_str, urls)."""
context_parts = []