fix(ai): increase OpenRouter max_tokens to 16384

aaronsb · aaronsb · commit d9c6e43d6068 · 2026-03-15T17:02:37.000-05:00
4096 truncates extraction responses for verbose models via OpenRouter.
Bump to 16384 as interim fix — proper solution is catalog-driven token
limits per model.
diff --git a/api/app/lib/ai_providers.py b/api/app/lib/ai_providers.py
@@ -1028,7 +1028,7 @@ def extract_concepts(
                     {"role": "system", "content": system_prompt},
                     {"role": "user", "content": f"Text to analyze:\n\n{text}"},
                 ],
-                max_tokens=4096,
+                max_tokens=16384,
                 temperature=0.3,
                 response_format={"type": "json_object"},
             )

Original file line number	Diff line number	Diff line change
`@@ -1028,7 +1028,7 @@ def extract_concepts(`
`1028`	`1028`	`{"role": "system", "content": system_prompt},`
`1029`	`1029`	`{"role": "user", "content": f"Text to analyze:\n\n{text}"},`
`1030`	`1030`	`],`
`1031`		`- max_tokens=4096,`
	`1031`	`+ max_tokens=16384,`
`1032`	`1032`	`temperature=0.3,`
`1033`	`1033`	`response_format={"type": "json_object"},`
`1034`	`1034`	`)`