From 769e0d4746b9792f6dbf78ee423dc5167647f525 Mon Sep 17 00:00:00 2001
From: Yiorgis Gozadinos <ggozadinos@gmail.com>
Date: Mon, 5 Jan 2026 14:33:58 +0200
Subject: [PATCH] Update picture description prompt, move its config under
 prompts

---
 docs/configuration/processing.md                | 17 +++++++++++++++--
 haiku_rag_slim/haiku/rag/config/models.py       |  7 ++++++-
 .../haiku/rag/converters/docling_local.py       |  4 +++-
 tests/test_converters.py                        |  6 ++----
 4 files changed, 26 insertions(+), 8 deletions(-)

diff --git a/docs/configuration/processing.md b/docs/configuration/processing.md
index 17d18da8..9d88f5e0 100644
--- a/docs/configuration/processing.md
+++ b/docs/configuration/processing.md
@@ -99,7 +99,6 @@ conversion_options:
     model:
       provider: ollama             # ollama, openai, or custom
       name: ministral-3            # VLM model name
-    prompt: "Describe this image in detail. Be precise and concise."
     timeout: 90                    # Request timeout in seconds
     max_tokens: 200                # Maximum tokens in response
 ```
@@ -111,10 +110,24 @@ conversion_options:
   - `provider`: `ollama` (default), `openai`, or use `base_url` for custom endpoints
   - `name`: Model name (e.g., `ministral-3`, `granite3.2-vision`, `gpt-4-vision`)
   - `base_url`: Optional custom API endpoint for vLLM, LM Studio, etc.
-- **prompt**: Instruction for the VLM when describing images
 - **timeout**: Request timeout in seconds
 - **max_tokens**: Maximum tokens in the VLM response
 
+**Default prompt** (configured in `prompts.picture_description`):
+
+```
+Describe this image for a blind user. State the image type
+(screenshot, chart, photo, etc.), what it depicts, any visible text,
+and key visual details. Be concise and accurate.
+```
+
+To customize the prompt globally:
+
+```yaml
+prompts:
+  picture_description: "Your custom prompt here..."
+```
+
 **Using with Ollama:**
 
 ```yaml
diff --git a/haiku_rag_slim/haiku/rag/config/models.py b/haiku_rag_slim/haiku/rag/config/models.py
index 1b2f32be..dcc58f2b 100644
--- a/haiku_rag_slim/haiku/rag/config/models.py
+++ b/haiku_rag_slim/haiku/rag/config/models.py
@@ -106,7 +106,6 @@ class PictureDescriptionConfig(BaseModel):
             name="ministral-3",
         )
     )
-    prompt: str = "Describe this image in detail. Be precise and concise."
     timeout: int = 90
     max_tokens: int = 200
 
@@ -186,6 +185,12 @@ class PromptsConfig(BaseModel):
     domain_preamble: str = ""
     qa: str | None = None
     synthesis: str | None = None
+    picture_description: str = (
+        "Describe this image for a blind user. "
+        "State the image type (screenshot, chart, photo, etc.), "
+        "what it depicts, any visible text, and key visual details. "
+        "Be concise and accurate."
+    )
 
 
 class AppConfig(BaseModel):
diff --git a/haiku_rag_slim/haiku/rag/converters/docling_local.py b/haiku_rag_slim/haiku/rag/converters/docling_local.py
index 016b6200..28305279 100644
--- a/haiku_rag_slim/haiku/rag/converters/docling_local.py
+++ b/haiku_rag_slim/haiku/rag/converters/docling_local.py
@@ -117,6 +117,8 @@ class DoclingLocalConverter(DocumentConverter):
         if pic_desc.enabled:
             from pydantic import AnyUrl
 
+            prompt = self.config.prompts.picture_description
+
             pipeline_options.enable_remote_services = True
             pipeline_options.picture_description_options = PictureDescriptionApiOptions(
                 url=AnyUrl(self._get_vlm_api_url(pic_desc.model)),
@@ -124,7 +126,7 @@ class DoclingLocalConverter(DocumentConverter):
                     model=pic_desc.model.name,
                     max_completion_tokens=pic_desc.max_tokens,
                 ),
-                prompt=pic_desc.prompt,
+                prompt=prompt,
                 timeout=pic_desc.timeout,
             )
 
diff --git a/tests/test_converters.py b/tests/test_converters.py
index 63c03f04..0ff9d48d 100644
--- a/tests/test_converters.py
+++ b/tests/test_converters.py
@@ -361,19 +361,17 @@ class TestDoclingLocalConverter:
         assert (
             config.processing.conversion_options.picture_description.max_tokens == 200
         )
+        # Default prompt is in PromptsConfig
+        assert "blind user" in config.prompts.picture_description
 
     def test_picture_description_config_applied(self, config):
         """Test that picture description config is applied to converter."""
         config.processing.conversion_options.picture_description.enabled = True
-        config.processing.conversion_options.picture_description.prompt = (
-            "Custom prompt for testing."
-        )
         config.processing.conversion_options.picture_description.timeout = 120
         converter = DoclingLocalConverter(config)
 
         pic_desc = converter.config.processing.conversion_options.picture_description
         assert pic_desc.enabled is True
-        assert pic_desc.prompt == "Custom prompt for testing."
         assert pic_desc.timeout == 120