llm_ticket3/prompts/test_prompt_ocr2.txt

params = {
            "stream": False,
            "seed": 0,
            #"stop_sequence": [],
            "temperature": 1.5,
            #"reasoning_effort": 0.5,
            #"logit_bias": {},
            "mirostat": 0,
            "mirostat_eta": 0.1,
            "mirostat_tau": 5.0,
            "top_k": 40,
            "top_p": 0.85,
            "min_p": 0.05,
            "frequency_penalty": 0.0,
            "presence_penalty": 0.0,
            "repeat_penalty": 1.1,
            "repeat_last_n": 128,
            "tfs_z": 1.0,
            "num_keep": 0,
            "num_predict": 4096,
            "num_ctx": 16384,
            #"repeat_penalty": 1.1,
            "num_batch": 2048,
            #"mmap": True,
            #"mlock": False,
            #"num_thread": 4,
            #"num_gpu": 1                   
        }

Essai prompt fusionné 1:

Your task is to perform a highly advanced OCR extraction from a technical webpage screenshot.

Extract **all** text content, regardless of size, clarity, or position on screen. Be exhaustive and structured.

STRUCTURE YOUR OUTPUT AS FOLLOWS:

1. HEADER INFORMATION
* Document title or test name
* URL and standard references
* Date stamps and user/author names

2. INTERFACE ELEMENTS
* Navigation menus, buttons, tabs
* Sidebar content and labels
* Form fields or table headers

3. CONTENT TEXT
* Any test names, materials, or results shown
* Any sample IDs, values, technical text

4. SYSTEM AND ERROR MESSAGES
* Server or status messages (e.g., “Impossible de trouver l'adresse”)
* Domain, IP, status/error codes

5. METADATA
* Reference numbers
* Version info, timestamps
* Operator or laboratory identifiers

6. PERIPHERAL OR FAINT TEXT
* Watermarks, logos, background overlays
* Faint symbols, truncated/partial lines (use “[...]” if incomplete)

RULES:
- Keep original formatting and spelling (no corrections or interpretations)
- Include duplicated elements
- Do not paraphrase
- Preserve punctuation, accents, capitalization

Mark each category clearly and output in clean bullet-point format.

Résultats params par défaut:
🔵 Résultat pour l'image output/ticket_T11143/T11143_20250422_084617/attachments/image_145435.png:
Texte extrait:
The webpage appears to be a technical document or report written entirely in French, which could make extraction challenging due to language barriers.

**HEADER INFORMATION**

No visible header information was found on the provided screenshot.

**INTERFACE ELEMENTS**

Navigation menus, buttons, tabs:

    • A drop-down menu with two options: "Matiériau" and "Observations."
    • Four checkboxes next to "MATERIEL," "PORTFOLIO," "OBSERVATIONS," and "SMQ."

Sidebar content and labels:

    • The text reads, "Bollee 1: CARRIÈRE ADÉG."

Form fields or table headers:

    • "Chantillion n°: 25-00075 receptionnée le 02/04/2025 par BOLLÉE Victor - prélèvée le 02/04/2025 par BOLLÉE Victor."
    • "Matériau: Sable 0/2 C - CARRIÈRE ADÉG."

**CONTENT TEXT**

Any test names, materials, or results shown:

    • Essai au bleu de méthylène (MB)
    • NF EN 933-9 (02-2007) - Essai de détermination du sable - Partie 9 : Dénombrement des particules

Sample IDs, values, technical text:

    • No sample ID information is visible.
    • "25-00075 réceptionnée le 02/04/2025 par BOLLÉE Victor - prélèvée le 02/04/2025 par BOLLÉE Victor."
    • "Sable 0/2 C - CARRIÈRE ADÉG"

**SYSTEM AND ERROR MESSAGES**

Server or status messages:

    • None visible.

Domain, IP, status/error codes:

    • No information available in the screenshot provided.

Résultats params:
{
  "temperature": 1.4,
  "top_p": 0.85,
  "frequency_penalty": 0.1,
  "presence_penalty": 0.1,
  "repeat_penalty": 1.05,
  "top_k": 30,
  "min_p": 0.08
}
🔵 Résultat pour l'image output/ticket_T11143/T11143_20250422_084617/attachments/image_145435.png:
Texte extrait:
The following screenshot presents a technical webpage featuring a table with text content and system messages. The output is organized into categories to ensure thoroughness and structure.

**1. HEADER INFORMATION**

*   **Document title or test name:**
    *   Essai au bleu de méthylène (MB) - NF EN 933-9 (02-2022)

**2. INTERFACE ELEMENTS**

*   **Navigation menus, buttons, tabs:** None visible
*   **Sidebar content and labels:**
    *   Matériau
    *   Table 0/2.C - CARRIÈRE ADCEG
*   **Form fields or table headers:** None visible

**3. CONTENT TEXT**

*   **Any test names, materials, or results shown:**
    *   Essai au bleu de méthylène (MB) - NF EN 933-9 (02-2022)
    *   Matériau: Table 0/2.C - CARRIÈRE ADCEG

**4. SYSTEM AND ERROR MESSAGES**

*   **Server or status messages:** Impossible de trouver l'adresse

**5. METADATA**

*   **Reference numbers:** None visible
*   **Version info, timestamps:** None visible
*   **Operator or laboratory identifiers:** None visible

**6. PERIPHERAL OR FAINT TEXT**

*   **Watermarks, logos, background overlays:** BRG-LAB logo
*   **Faint symbols, truncated/partial lines:**
    *   \[...]

This analysis provides a detailed and structured breakdown of the text content in the screenshot.

----


Prompt fusioné 2:
Scan the image in a structured grid from top-left to bottom-right, capturing all visible text with administrative precision.

For each detected element, report:

- POSITION: (Top/Center/Bottom + Left/Right)
- TEXT: (Exact content)
- TYPE: Printed / Handwritten / Error / Label / Button
- CONTEXT: If part of a form, menu, or system message

MANDATORY ELEMENTS TO EXTRACT:
* Document titles, reference standards (NF EN etc.)
* UI labels like “RAZ”, “IMPRIMER”, “MATÉRIEL”
* Sample IDs or values (e.g., “Sable 0/2 C – CARRIÈRE ADCEG”)
* Server/IP/domain errors
* Any visible metadata or date

Be exhaustive. Include:
- Truncated or cropped text
- Faint, partially displayed UI elements
- Sidebar/menu contents
- Stamps or logos

Respect original formatting (use “[...]” where needed). No paraphrasing.

Résultats params par défaut:
🔵 Résultat pour l'image output/ticket_T11143/T11143_20250422_084617/attachments/image_145435.png:
Texte extrait:
The image displays a screenshot of a web page with French text. The top-left corner has a blue bar with white text that reads "BRG-LAB". Below this is a gray box with the title "Essai au bleu de méthylène (MB) - NF EN 933-9" in black text.

To the left of the title are several buttons and links:

* A button labeled "RAZ"
* A button labeled "IMPRIMER"
* A link labeled "MATÉRIEL"

Below these buttons and links is a table with several columns:

* A column labeled "SABLE 0/2 C - CARRIÈRE ADCEG"
* A column labeled "DATE DE RÉCEPTION"
* A column labeled "RÉSULTAT"

In the bottom-right corner, there is a gray box with white text that reads "Impossible de trouver l'adresse IP du serveur de zk1.brg-lab.com." 

The background of the image is a light beige color. Overall, the image appears to be a screenshot of a web page used for managing and tracking laboratory tests or experiments.

Résultats params défaut sauf num_predict: 2048:
🔵 Résultat pour l'image output/ticket_T11143/T11143_20250422_084617/attachments/image_145435.png:
Texte extrait:
The image displays a screenshot of a web page with French text. The top-left corner has a blue logo that says "BRG-LAB" in white letters, and the title bar reads "Essai au bleu de méthylène (MB) - NF EN 933-9 (02-2022)".

**Text Elements:**

* **Title:** Essai au bleu de méthylène (MB) - NF EN 933-9 (02-2022)
* **Logo:** BRG-LAB
* **Menu Items:**
        + RAZ
        + IMPRIMER
        + MATÉRIEL
* **Sample ID/Value:** Sable 0/2 C – CARRIÈRE ADCEG
* **Error Message:** Impossible de trouver l'adresse IP du serveur de zik1.brg-lab.com

**Other Elements:**

* A blue sidebar on the left side of the page with a list of menu items.
* A search bar at the top of the page.
* A table or grid displaying data related to the title.

**Positioning and Formatting:**

* The title is centered at the top of the page.
* The logo is located in the top-left corner.
* The menu items are listed vertically in the blue sidebar.
* The sample ID/value is displayed below the title.
* The error message is displayed at the bottom of the page.

Résultats params:
{
  "temperature": 1.6,
  "top_p": 0.88,
  "repeat_penalty": 1.15,
  "top_k": 45,
  "presence_penalty": 0.0
}
🔵 Résultat pour l'image output/ticket_T11143/T11143_20250422_084617/attachments/image_145435.png:
Texte extrait:
The image displays a screenshot of the Giraud Brg-Lab website, which appears to be a French-language interface for managing and analyzing materials science data. The top section contains the page title "Essai au bleu de méthylène" with a reference standard "(MB) - NF EN 933-9". Below this is a navigation bar featuring buttons such as "R.A.Z.", "IMPRIMER", and "MATÉRIEL".

In the main content area, there are several fields labeled "ESSAI", "MATERIEL", "PORTFOLIO", "OBSERVATIONS", "SMQ", and "HISTORIQUE". These labels likely correspond to different types of data related to the material being analyzed.

At the bottom of the page, there is a copyright notice that reads "© Giraud Brg-Lab SA" followed by the year "2022". Additionally, there is an icon representing a server or IP address error with text stating "Erreur de serveur (IP)" above it. Next to this icon, another error message displays "Erreur d'accès à la base de données".

On the left side of the page, there are three boxes containing information about the user's account. The first box shows the username "BOLLEE Victor", while the second and third boxes display the dates "02/04/2025 14:17:22 CEST" and "12/31/9999", respectively.

Overall, the image suggests that the Giraud Brg-Lab website provides tools for scientists to record and analyze data related to materials science experiments. However, due to errors present on the page, some functionality may be impaired.

Résultats params:
{
  "temperature": 1.6,
  "top_p": 0.88,
  "repeat_penalty": 1.15,
  "top_k": 45,
  "presence_penalty": 0.0
  "num_predict": 2048 
}
🔵 Résultat pour l'image output/ticket_T11143/T11143_20250422_084617/attachments/image_145435.png:
Texte extrait:
The image displays a screenshot of the Giraud Brg-Lab website, which appears to be a French-language interface for managing and analyzing materials science data. The top section contains the page title "Essai au bleu de méthylène" with a reference standard "(MB) - NF EN 933-9". Below this is a navigation bar featuring buttons such as "R.A.Z.", "IMPRIMER", and "MATÉRIEL".

In the main content area, there are several fields labeled "ESSAI", "MATERIEL", "PORTFOLIO", "OBSERVATIONS", "SMQ", and "HISTORIQUE". These labels likely correspond to different types of data related to the material being analyzed.

At the bottom of the page, there is a copyright notice that reads "© Giraud Brg-Lab SA" followed by the year "2022". Additionally, there is an icon representing a server or IP address error with text stating "Erreur de serveur (IP)" above it. Next to this icon, another error message displays "Erreur d'accès à la base de données".

On the left side of the page, there are three boxes containing information about the user's account. The first box shows the username "BOLLEE Victor", while the second and third boxes display the dates "02/04/2025 14:17:22 CEST" and "12/31/9999", respectively.

Overall, the image suggests that the Giraud Brg-Lab website provides tools for scientists to record and analyze data related to materials science experiments. However, due to errors present on the page, some functionality may be impaired.


Prompt fusionné 3:
Extract all visible and invisible (background, faded, small) text from a technical web UI screenshot.

CATEGORIES TO INCLUDE:
1. PAGE STRUCTURE
* Page title, sections, headers, subtitles

2. LABORATORY/TECHNICAL CONTENT
* Standard references (e.g., NF EN 933-9), test names
* Sample codes, material types
* Any observed values or test metadata

3. UI TEXT
* Buttons: (RAZ, IMPRIMER, ENREGISTRER, etc.)
* Tabs and navigation: (MATÉRIEL, PORTFOLIO, etc.)
* Labels or side panel content

4. ERRORS/NOTIFICATIONS
* System messages (e.g., IP not found, connection failures)
* Domain/server names (e.g., zkt1.brg-lab.com)

RULES:
- Keep exact format and line breaks
- Preserve upper/lower case
- Include duplicated content
- Use “[...]” for truncated or unclear text

FORMAT:
Output as bullet lists grouped by category. Do not interpret or translate. Output in original language.

Résultats params par défaut:
🔵 Résultat pour l'image output/ticket_T11143/T11143_20250422_084617/attachments/image_145435.png:
Texte extrait:
Here are the contents of your image categorized into four groups:

**PAGE STRUCTURE**

• Essai au bleu de méthylène (MB) - NF EN 933-9
• BOLLEE Victor 
• 25/08/2025 réceptionné le 02/04/2025

**LABORATORY/TECHNICAL CONTENT**

• Standard reference: NF EN 933-9
• Test name: Essai au bleu de méthylène (MB)
• Sample code/material type: Not explicitly stated, but implies a material related to "Bollee Victor" based on context
• Observed values/test metadata: Not provided in the visible screenshot; potentially contained within the "[...]" truncated or unclear content sections

**UI TEXT**

• Buttons:
        + RAZ (clear/reset)
        + IMPRIMER (print)
        + ENREGISTRER (register/save)

**ERRORS/NOTIFICATIONS**

• Error message: "[IP du serveur de zkt1.brg-lab.com.]"
Note that some details are obscured by the redaction of parts of the URL, preventing further interpretation without additional information.

Résultat params par défaut suf num_predict: 2048:
🔵 Résultat pour l'image output/ticket_T11143/T11143_20250422_084617/attachments/image_145435.png:
Texte extrait:
Here are the categories you requested:

**PAGE STRUCTURE**

• Essai au bleu de méthylène (MB) - NF EN 933-9 (02-2025)
• Echantillon : n° 25-00075 réceptionné le 02/04/2025 par BOLLEE Victor - prélèvement le 02/04/2025 par BOLLEE Victor, n° prélèvement : 25-00075
• Matériau : Sable 0/2 C - CARREIRE ADCEG

**LABORATORY/TECHNICAL CONTENT**

• Standard reference: NF EN 933-9
• Test name: Essai au bleu de méthylène (MB)
• Sample code: 25-00075
• Material type: Sable 0/2 C
• Observed values:
        + Date of reception: 02/04/2025
        + Date of sampling: 02/04/2025

**UI TEXT**

• Buttons:
        + RAZ
        + IMPRIMER
        + ENREGISTRER
• Tabs and navigation:
        + MATÉRIEL
        + PORTFOLIO
        + OBSERVATIONS
        + SMO
        + HISTORIQUE

**ERRORS/NOTIFICATIONS**

• None visible

Note: There are no errors or notifications visible on the page.

Résultat params:
{
  "temperature": 1.3,
  "top_p": 0.8,
  "repeat_penalty": 1.2,
  "frequency_penalty": 0.15,
  "top_k": 35,
  "min_p": 0.06
}

Résultat params:
{
  "temperature": 1.3,
  "top_p": 0.8,
  "repeat_penalty": 1.2,
  "frequency_penalty": 0.15,
  "top_k": 35,
  "min_p": 0.06,
  "num_predict": 2048
}

🔵 Résultat pour l'image output/ticket_T11143/T11143_20250422_084617/attachments/image_145435.png:
Texte extrait:
Here are the requested technical UI screenshot elements:

**PAGE STRUCTURE**

• Essai au bleu de méthylène (MB) - NF EN 933-9 
    * Date: 02/04/2025  
    • Victor 

**LABORATORY / TECHNICAL CONTENT**
No laboratory references found.

* Sample codes, material types: None visible/invisible data detected.



UI TEXT

Tabs and navigation options include:

1. MATÉRIEL
2. PORTFOLIO
3. OBSERVATIONS



Buttons present on page:
RAZ, IMPRIMER


ERRORS / NOTIFICATIONS

An IP server error was noted during connection attempt to zkt1.brg-lab.com

Résultat params:
{
  "temperature": 1,
  "top_p": 0.8,
  "repeat_penalty": 1.2,
  "frequency_penalty": 0.15,
  "top_k": 35,
  "min_p": 0.06,
  "num_predict": 2048
}
🔵 Résultat pour l'image output/ticket_T11143/T11143_20250422_084617/attachments/image_145435.png:
Texte extrait:
Here are the results:

**PAGE STRUCTURE**
• Essai au bleu de méthylène (MB) - NF EN 933-9 (02-2022)

**LABORATORY/TECHNICAL CONTENT**

• Echantillon: n° 25-00075 réceptionné le 02/04/2025 par BOLLEE Victor
• Matériau: Sable d72 C - Carrière ADCEG

**UI TEXT**

• RAZ
• IMPRIMER
• ENREGISTRER
• MATÉRIEL
• PORTFOLIO
• OBSERVATIONS
• SMO
• HISTORIQUE

**ERRORS/NOTIFICATIONS**
None found.

Please note that the screenshot is mostly blank, so there isn't much text to extract. The "Impossibilité de trouver l'adresse IP du serveur de zkt1.brg-lab.com" error message at the bottom of the page was not included in the initial list as it appears outside of any specific category and might be considered a server-side issue rather than an element directly visible on this UI screenshot alone.

Résultat params:
{
  "temperature": 0.5,
  "top_p": 0.8,
  "repeat_penalty": 1.2,
  "frequency_penalty": 0.15,
  "top_k": 35,
  "min_p": 0.06,
  "num_predict": 2048
}
🔵 Résultat pour l'image output/ticket_T11143/T11143_20250422_084617/attachments/image_145435.png:
Texte extrait:
**PAGE STRUCTURE**

• Essai au bleu de méthylène (MB) - NF EN 933-9 
• BRG-LAB 

**LABORATORY/TECHNICAL CONTENT**

• Echantillon n°: 25-00075 réceptionné le 02/04/2025 par BOLLEE Victor  
• Matériel: Sable d72 C - Carrière ADCEQ  

**UI TEXT**

• RAZ
• IMPRIMER
• ENREGISTRER
• MATÉRIEL
• PORTFOLIO
• OBSERVATIONS
• SMO

**ERRORS/NOTIFICATIONS**

No error messages found.

Résultat params:
{
  "temperature": 0.5,
  "top_p": 0.8,
  "repeat_penalty": 1.2,
  "frequency_penalty": 0.15,
  "top_k": 35,
  "min_p": 0.06,
  "num_predict": 4096
}

🔵 Résultat pour l'image output/ticket_T11143/T11143_20250422_084617/attachments/image_145435.png:
Texte extrait:
**PAGE STRUCTURE**

• Essai au bleu de méthylène (MB) - NF EN 933-9 
• BRG-LAB 

**LABORATORY/TECHNICAL CONTENT**

• Echantillon n°: 25-00075 réceptionné le 02/04/2025 par BOLLEE Victor  
• Matériel: Sable d72 C - Carrière ADCEQ  

**UI TEXT**

• RAZ
• IMPRIMER
• ENREGISTRER
• MATÉRIEL
• PORTFOLIO
• OBSERVATIONS
• SMO

**ERRORS/NOTIFICATIONS**
No errors or notifications found.