patch of feat: add keep_alive header to API requests for improved connection management · round #0 · pull #1 · knbnnot.bsky.social/unvoid

+4 -4

ollama_adapter.py

···

       114
       114
        
                   data.update(filtered)

     

       115
       115
        
       

     

       116
       116
        
               body = json.dumps(data).encode('utf-8')

     

       117
       117
       -
               req = Request(self._url(generic_endpoint), data=body, headers={'Content-Type': 'application/json'})

     

       117
       117
       +
               req = Request(self._url(generic_endpoint), data=body, headers={'Content-Type': 'application/json', 'keep_alive': '180'})

     

       118
       118
        
               try:

     

       119
       119
        
                   with urlopen(req, timeout=self.timeout) as resp:

     

       120
       120
        
                       raw = resp.read().decode('utf-8')

     
···

       296
       296
        
                           logger.debug('Ollama embeddings request to %s; body=%s', self._url(ep), json.dumps(data)[:2000])

     

       297
       297
        
                       except Exception:

     

       298
       298
        
                           logger.debug('Ollama embeddings request prepared')

     

       299
       299
       -
                       req = Request(self._url(ep), data=body, headers={'Content-Type': 'application/json'})

     

       299
       299
       +
                       req = Request(self._url(ep), data=body, headers={'Content-Type': 'application/json', 'keep_alive': '180'})

     

       300
       300
        
                       try:

     

       301
       301
        
                           with urlopen(req, timeout=self.timeout) as resp:

     

       302
       302
        
                               raw = resp.read().decode('utf-8')

     
···

       369
       369
        
               # Try /api/tags (some Ollama builds) then /api/list

     

       370
       370
        
               for path in ['/api/tags', '/api/list']:

     

       371
       371
        
                   try:

     

       372
       372
       -
                       req = Request(self._url(path))

     

       372
       372
       +
                       req = Request(self._url(path), headers={'keep_alive': '180'})

     

       373
       373
        
                       with urlopen(req, timeout=5) as resp:

     

       374
       374
        
                           raw = resp.read().decode('utf-8')

     

       375
       375
        
                           try:

     
···

       409
       409
        
           )

     

       410
       410
        
           # quick health check (non-raising)

     

       411
       411
        
           try:

     

       412
       412
       -
               req = Request(client._url('/api/list'))

     

       412
       412
       +
               req = Request(client._url('/api/list'), headers={'keep_alive': '180'})

     

       413
       413
        
               with urlopen(req, timeout=2):

     

       414
       414
        
                   # ignore content; just successful connect

     

       415
       415
        
                   pass

+2 -2

ollama_introspection.py

···

       57
       57
        
       

     

       58
       58
        
       def _request_json(path: str, base_url: Optional[str] = None, timeout: int = 10) -> Dict[str, Any]:

     

       59
       59
        
           full = urljoin(_base_url(base_url) + '/', path.lstrip('/'))

     

       60
       60
       -
           req = Request(full)

     

       60
       60
       +
           req = Request(full, headers={'keep_alive': '180'})

     

       61
       61
        
           try:

     

       62
       62
        
               with urlopen(req, timeout=timeout) as resp:

     

       63
       63
        
                   raw = resp.read().decode('utf-8')

     
···

       124
       124
        
           path = '/api/show'

     

       125
       125
        
           full = urljoin(_base_url(base_url) + '/', path.lstrip('/'))

     

       126
       126
        
           body = json.dumps(payload).encode('utf-8')

     

       127
       127
       -
           req = Request(full, data=body, headers={'Content-Type': 'application/json'})

     

       127
       127
       +
           req = Request(full, data=body, headers={'Content-Type': 'application/json', 'keep_alive': '180'})

     

       128
       128
        
           try:

     

       129
       129
        
               with urlopen(req, timeout=10) as resp:

     

       130
       130
        
                   raw = resp.read().decode('utf-8')

+2 -2

scripts/ollama_demo.py

···

       43
       43
        
                   cfg = get_ollama_config()

     

       44
       44
        
                   url = cfg.get('url').rstrip('/') + '/api/generate'

     

       45
       45
        
                   body = json.dumps({'model': client.chat_model, 'prompt': prompt}).encode('utf-8')

     

       46
       46
       -
                   req = Request(url, data=body, headers={'Content-Type': 'application/json'})

     

       46
       46
       +
                   req = Request(url, data=body, headers={'Content-Type': 'application/json', 'keep_alive': '180'})

     

       47
       47
        
                   with urlopen(req, timeout=5) as resp:

     

       48
       48
        
                       raw = resp.read().decode('utf-8')

     

       49
       49
        
                       print('\n=== Raw /api/generate response ===')

     
···

       70
       70
        
                   cfg = get_ollama_config()

     

       71
       71
        
                   url = cfg.get('url').rstrip('/') + f'/api/models/{client.embedding_model}/embed'

     

       72
       72
        
                   body = json.dumps({'input': texts}).encode('utf-8')

     

       73
       73
       -
                   req = Request(url, data=body, headers={'Content-Type': 'application/json'})

     

       73
       73
       +
                   req = Request(url, data=body, headers={'Content-Type': 'application/json', 'keep_alive': '180'})

     

       74
       74
        
                   with urlopen(req, timeout=5) as resp:

     

       75
       75
        
                       raw = resp.read().decode('utf-8')

     

       76
       76
        
                       print('\n=== Raw model-specific embed response ===')