janito 1.7.0__tar.gz → 1.9.0__tar.gz

janito-1.9.0/MANIFEST.in

@@ -0,0 +1,6 @@
+# Exclude all .bak files from the package
+exclude *.bak
+recursive-exclude * *.bak
+
+# Include static files for termweb
+recursive-include janito/termweb/static *.*

{janito-1.7.0/janito.egg-info → janito-1.9.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: janito
-Version: 1.7.0
+Version: 1.9.0
 Summary: Natural Language Coding Agent,
 Author-email: João Pinto <joao.pinto@gmail.com>
 License-Expression: MIT
@@ -22,6 +22,7 @@ Requires-Dist: prompt_toolkit
 Requires-Dist: requests
 Requires-Dist: rich
 Requires-Dist: toml
+Requires-Dist: lxml
 Provides-Extra: docs
 Requires-Dist: mkdocs>=1.5.3; extra == "docs"
 Requires-Dist: mkdocs-material>=9.4.8; extra == "docs"
@@ -29,19 +30,53 @@ Dynamic: license-file
 
 # 🚀 Janito: Natural Language Coding Agent
 
-**Current Version: 1.7.0**  
+**Current Version: 1.8.1**  
 See [CHANGELOG.md](./CHANGELOG.md) for details on the latest release.
 
 Janito is an AI-powered assistant for the command line and web that interprets natural language system_prompt_template to edit code, manage files, and analyze projects using patterns and tools designed by experienced software engineers. It prioritizes transparency, interactive clarification, and precise, reviewable changes.
 
 For a technical overview, see the [Architecture Guide](docs/reference/architecture.md).
 
+## 🧪 Experimental Feature: termweb File Viewer
+
+Janito now includes an experimental lightweight web file viewer for use with the CLI chat shell. This feature allows you to click on file paths in the terminal (when using a Rich-compatible terminal) and instantly preview file contents in your browser—no full IDE required!
+
+### How to Use
+
+- Start the CLI chat shell with the `--termweb` flag:
+  ```bash
+  janito --termweb
+  ```
+- To disable the termweb file viewer, use the `--no-termweb` flag.
+- By default, the viewer runs at http://localhost:8088 (or the next available port up to 8100).
+- To specify a port, use `--termweb-port 8090`.
+- File paths in CLI output become clickable links that open the file in your browser.
+
+**Note:** This feature is experimental. It is intended for quick file previews and review, not for editing or production use. Feedback is welcome!
+
+### Why is this useful?
+- Enables instant file previews from the CLI without a full IDE.
+- Works with all Janito file tools and outputs that display file paths.
+- Uses the Rich library’s link markup for clickable terminal links.
+
+---
+
 ## 📖 Full Documentation & Overview
 - For structured and in-depth guides, visit the [Janito Documentation Site](https://docs.janito.dev).
 - For a high-level, user-friendly overview, see [janito.dev](https://janito.dev).
 
 ---
 
+## Listing Available Tools
+
+To list all registered tools on the command line, use the option:
+
+```sh
+janito --list-tools
+```
+
+This will display a colorful table with the name, description, and parameters of each available tool.
+
 ## ⚡ Quick Start
 
 ## 🖥️ Supported Human Interfaces
@@ -61,6 +96,7 @@ You can alter Janito's behavior in any interface using these flags:
 - `--no-tools`: Disable all tool usage (Janito will only use the language model, no file/code/shell actions).
 - `--trust-tools`/`-T`: Trusted tools mode (suppresses all tool output, only shows output file locations).
 - `--vanilla`: Disables tools, system prompt, and temperature settings for a pure LLM chat experience.
+- `--no-termweb`: Disables the termweb file viewer for terminal links.
 
 These modifiers can be combined with any interface mode for tailored workflows.
 
@@ -89,7 +125,7 @@ janito --web
 - 📁 **File & Directory Management:** Navigate, create, move, or remove files and folders.
 - 🧠 **Context-Aware:** Understands your project structure for precise edits.
 - 💬 **Interactive User Prompts:** Asks for clarification when needed.
-- 🧩 **Extensible Tooling:** Built-in tools for file operations, shell commands, directory and file management, Python code execution and validation, text replacement, and more.
+- 🛠️ **Extensible Tooling:** Built-in tools for file operations, shell commands, directory and file management, Python code execution and validation, text replacement, and more.
   - See [janito/agent/tools/README.md](janito/agent/tools/README.md) for the full list of built-in tools and their usage details. For the message handler model, see [docs/MESSAGE_HANDLER_MODEL.md](docs/MESSAGE_HANDLER_MODEL.md).
 - 🌐 **Web Interface (In Development):** Simple web UI for streaming responses and tool progress.
 
@@ -101,7 +137,7 @@ janito --web
 
 ### Contributing & Developer Guide
 
-If you want to extend Janito or add new tools, see the [Developer Guide](docs/README_DEV.md) for system_prompt_template, tool registration requirements, and code style guidelines. For the full list of built-in tools and their usage, see the [Tools Reference](janito/agent/tools/README.md).
+If you want to extend Janito or add new tools, see the [Developer Guide](docs/README_DEV.md) for system_prompt_template, tool registration requirements, and code profile guidelines. For the full list of built-in tools and their usage, see the [Tools Reference](janito/agent/tools/README.md).
 
 
 
@@ -138,12 +174,12 @@ For details on using models hosted on Azure OpenAI, see [docs/AZURE_OPENAI.md](d
 
 ---
 
-## 🧩 System Prompt & Role
+## 🧑‍💻 System Prompt & Role
 
-Janito operates using a system prompt template that defines its behavior, communication style, and capabilities. By default, Janito assumes the role of a "software engineer"—this means its responses and actions are tailored to the expectations and best practices of professional software engineering.
+Janito operates using a system prompt template that defines its behavior, communication profile, and capabilities. By default, Janito assumes the role of a "software engineer"—this means its responses and actions are tailored to the expectations and best practices of professional software engineering.
 
 - **Role:** You can customize the agent's role (e.g., "data scientist", "DevOps engineer") using the `--role` flag or config. The default is `software engineer`.
-- **System Prompt Template:** The system prompt is rendered from a Jinja2 template (see `janito/agent/templates/system_prompt_template.j2`). This template governs how the agent interprets system_prompt_template, interacts with files, and communicates with users.
+- **System Prompt Template:** The system prompt is rendered from a Jinja2 template (see `janito/agent/templates/prompt_prompt_template.j2`). This template governs how the agent interprets system_prompt_template, interacts with files, and communicates with users.
 - **Customization & Precedence:** Advanced users can override the system prompt with the `--system` flag (raw string), or point to a custom file using `--system-file`. The precedence is: `--system-file` > `--system`/config > default template.
 
 The default template ensures the agent:
@@ -152,7 +188,7 @@ The default template ensures the agent:
 - Provides concise plans before taking action
 - Documents any changes made
 
-For more details or to customize the prompt, see the template file at `janito/agent/templates/system_prompt_template.j2` and the architecture overview in [docs/reference/architecture.md](docs/reference/architecture.md).
+For more details or to customize the prompt, see the template file at `janito/agent/templates/prompt_prompt_template.j2` and the architecture overview in [docs/reference/architecture.md](docs/reference/architecture.md).
 
 ---
 
@@ -184,18 +220,18 @@ Vanilla mode is ideal for:
 
 > Note: Vanilla mode is a runtime switch and does not change the Agent API or class signatures. It is controlled via CLI/config only.
 
-## 🧑‍💻 AgentProfileManager: Profile, Role, and Prompt Management
+## 👨‍💻 AgentProfileManager: Profile, Role, and Prompt Management
 
-Janito now uses a dedicated `AgentProfileManager` class to manage user profiles, roles, interaction styles, and system prompt selection. This manager:
-- Stores the current role (e.g., "software engineer") and interaction style (e.g., "default", "technical").
-- Renders the system prompt from the appropriate template based on interaction style.
+Janito now uses a dedicated `AgentProfileManager` class to manage user profiles, roles, interaction profiles, and system prompt selection. This manager:
+- Stores the current role (e.g., "software engineer") and interaction profile (e.g., "default", "technical").
+- Renders the system prompt from the appropriate template based on interaction profile.
 - Instantiates and manages the low-level LLM Agent, passing the correct prompt.
-- Provides methods to update the role, interaction style, and refresh the prompt at runtime.
+- Provides methods to update the role, interaction profile, and refresh the prompt at runtime.
 
 ### Multiple System Prompt Templates
-- The system prompt template is now selected based on the interaction style (e.g., `default` or `technical`).
+- The system prompt template is now selected based on the interaction profile (e.g., `default` or `technical`).
 - Templates are located in `janito/agent/templates/` (see `system_prompt_template.j2` and `system_prompt_template_technical.j2`).
-- You can switch interaction styles at runtime using the profile manager, enabling different agent behaviors for different user needs.
+- You can switch interaction profiles at runtime using the profile manager, enabling different agent behaviors for different user needs.
 
 This separation ensures that the LLM Agent remains focused on language model interaction and tool execution, while all profile, role, and prompt logic is managed at a higher level.
 
@@ -203,45 +239,45 @@ See `janito/agent/profile_manager.py` for implementation details.
 
 ### Agent Interaction Style
 
-You can control the agent's behavior and prompt style globally or per-project using the `interaction_style` config key. Supported values:
+You can control the agent's behavior and prompt profile globally or per-project using the `profile` config key. See [Prompt Profiles Guide](docs/guides/prompt_profiles.md) for all available styles and combinations.
 - `default`: Concise, general-purpose agent (default)
 - `technical`: Strict, workflow-oriented for technical/developer use
 
 Set globally:
 ```bash
-janito --set-global-config interaction_style=technical
+janito --set-global-config profile=technical
 ```
 Or per-project (in your project root):
 ```bash
-janito --set-local-config interaction_style=technical
+janito --set-local-config profile=technical
 ```
 
 You can also override for a session with the CLI flag:
 ```bash
-janito --style technical
+janito --profile technical
 ```
 
 See [docs/CONFIGURATION.md](docs/CONFIGURATION.md) for full details.
 
-## 🧩 Combinatorial Style System
+## 🧑‍💻 Combinatorial Style System
 
-Janito now supports combinatorial styles for system prompts, allowing you to combine a main style (such as `default` or `technical`) with one or more feature extensions (such as `commit_all`).
+Janito now supports combinatorial profiles for system prompts, allowing you to combine a main profile (such as `default` or `technical`) with one or more feature extensions (such as `commit_all`).
 
-- **Main style:** The base agent behavior and workflow (e.g., `default`, `technical`).
-- **Feature extensions:** Optional features that override or extend the main style (e.g., `commit_all`).
+- **Main profile:** The base agent behavior and workflow (e.g., `default`, `technical`).
+- **Feature extensions:** Optional features that override or extend the main profile (e.g., `commit_all`).
 - **Syntax:** Use a hyphen to combine, e.g., `technical-commit_all`.
 
 **How it works:**
-- The main style template is loaded first.
+- The main profile template is loaded first.
 - Each feature extension template is layered on top, overriding or extending specific blocks in the main template.
 - Feature templates must use `{% extends parent_template %}` for dynamic inheritance.
 
 **Example usage:**
 ```bash
-janito --style technical-commit_all
+janito --profile technical-commit_all
 ```
 
-This will apply the `technical` style with the `commit_all` feature enabled in the agent's system prompt.
+This will apply the `technical` profile with the `commit_all` feature enabled in the agent's system prompt.
 
 See `janito/render_prompt.py` and `janito/agent/templates/` for implementation details and to create your own feature extensions.
 

{janito-1.7.0 → janito-1.9.0}/README.md

@@ -1,18 +1,52 @@
 # 🚀 Janito: Natural Language Coding Agent
 
-**Current Version: 1.7.0**  
+**Current Version: 1.8.1**  
 See [CHANGELOG.md](./CHANGELOG.md) for details on the latest release.
 
 Janito is an AI-powered assistant for the command line and web that interprets natural language system_prompt_template to edit code, manage files, and analyze projects using patterns and tools designed by experienced software engineers. It prioritizes transparency, interactive clarification, and precise, reviewable changes.
 
 For a technical overview, see the [Architecture Guide](docs/reference/architecture.md).
 
+## 🧪 Experimental Feature: termweb File Viewer
+
+Janito now includes an experimental lightweight web file viewer for use with the CLI chat shell. This feature allows you to click on file paths in the terminal (when using a Rich-compatible terminal) and instantly preview file contents in your browser—no full IDE required!
+
+### How to Use
+
+- Start the CLI chat shell with the `--termweb` flag:
+  ```bash
+  janito --termweb
+  ```
+- To disable the termweb file viewer, use the `--no-termweb` flag.
+- By default, the viewer runs at http://localhost:8088 (or the next available port up to 8100).
+- To specify a port, use `--termweb-port 8090`.
+- File paths in CLI output become clickable links that open the file in your browser.
+
+**Note:** This feature is experimental. It is intended for quick file previews and review, not for editing or production use. Feedback is welcome!
+
+### Why is this useful?
+- Enables instant file previews from the CLI without a full IDE.
+- Works with all Janito file tools and outputs that display file paths.
+- Uses the Rich library’s link markup for clickable terminal links.
+
+---
+
 ## 📖 Full Documentation & Overview
 - For structured and in-depth guides, visit the [Janito Documentation Site](https://docs.janito.dev).
 - For a high-level, user-friendly overview, see [janito.dev](https://janito.dev).
 
 ---
 
+## Listing Available Tools
+
+To list all registered tools on the command line, use the option:
+
+```sh
+janito --list-tools
+```
+
+This will display a colorful table with the name, description, and parameters of each available tool.
+
 ## ⚡ Quick Start
 
 ## 🖥️ Supported Human Interfaces
@@ -32,6 +66,7 @@ You can alter Janito's behavior in any interface using these flags:
 - `--no-tools`: Disable all tool usage (Janito will only use the language model, no file/code/shell actions).
 - `--trust-tools`/`-T`: Trusted tools mode (suppresses all tool output, only shows output file locations).
 - `--vanilla`: Disables tools, system prompt, and temperature settings for a pure LLM chat experience.
+- `--no-termweb`: Disables the termweb file viewer for terminal links.
 
 These modifiers can be combined with any interface mode for tailored workflows.
 
@@ -60,7 +95,7 @@ janito --web
 - 📁 **File & Directory Management:** Navigate, create, move, or remove files and folders.
 - 🧠 **Context-Aware:** Understands your project structure for precise edits.
 - 💬 **Interactive User Prompts:** Asks for clarification when needed.
-- 🧩 **Extensible Tooling:** Built-in tools for file operations, shell commands, directory and file management, Python code execution and validation, text replacement, and more.
+- 🛠️ **Extensible Tooling:** Built-in tools for file operations, shell commands, directory and file management, Python code execution and validation, text replacement, and more.
   - See [janito/agent/tools/README.md](janito/agent/tools/README.md) for the full list of built-in tools and their usage details. For the message handler model, see [docs/MESSAGE_HANDLER_MODEL.md](docs/MESSAGE_HANDLER_MODEL.md).
 - 🌐 **Web Interface (In Development):** Simple web UI for streaming responses and tool progress.
 
@@ -72,7 +107,7 @@ janito --web
 
 ### Contributing & Developer Guide
 
-If you want to extend Janito or add new tools, see the [Developer Guide](docs/README_DEV.md) for system_prompt_template, tool registration requirements, and code style guidelines. For the full list of built-in tools and their usage, see the [Tools Reference](janito/agent/tools/README.md).
+If you want to extend Janito or add new tools, see the [Developer Guide](docs/README_DEV.md) for system_prompt_template, tool registration requirements, and code profile guidelines. For the full list of built-in tools and their usage, see the [Tools Reference](janito/agent/tools/README.md).
 
 
 
@@ -109,12 +144,12 @@ For details on using models hosted on Azure OpenAI, see [docs/AZURE_OPENAI.md](d
 
 ---
 
-## 🧩 System Prompt & Role
+## 🧑‍💻 System Prompt & Role
 
-Janito operates using a system prompt template that defines its behavior, communication style, and capabilities. By default, Janito assumes the role of a "software engineer"—this means its responses and actions are tailored to the expectations and best practices of professional software engineering.
+Janito operates using a system prompt template that defines its behavior, communication profile, and capabilities. By default, Janito assumes the role of a "software engineer"—this means its responses and actions are tailored to the expectations and best practices of professional software engineering.
 
 - **Role:** You can customize the agent's role (e.g., "data scientist", "DevOps engineer") using the `--role` flag or config. The default is `software engineer`.
-- **System Prompt Template:** The system prompt is rendered from a Jinja2 template (see `janito/agent/templates/system_prompt_template.j2`). This template governs how the agent interprets system_prompt_template, interacts with files, and communicates with users.
+- **System Prompt Template:** The system prompt is rendered from a Jinja2 template (see `janito/agent/templates/prompt_prompt_template.j2`). This template governs how the agent interprets system_prompt_template, interacts with files, and communicates with users.
 - **Customization & Precedence:** Advanced users can override the system prompt with the `--system` flag (raw string), or point to a custom file using `--system-file`. The precedence is: `--system-file` > `--system`/config > default template.
 
 The default template ensures the agent:
@@ -123,7 +158,7 @@ The default template ensures the agent:
 - Provides concise plans before taking action
 - Documents any changes made
 
-For more details or to customize the prompt, see the template file at `janito/agent/templates/system_prompt_template.j2` and the architecture overview in [docs/reference/architecture.md](docs/reference/architecture.md).
+For more details or to customize the prompt, see the template file at `janito/agent/templates/prompt_prompt_template.j2` and the architecture overview in [docs/reference/architecture.md](docs/reference/architecture.md).
 
 ---
 
@@ -155,18 +190,18 @@ Vanilla mode is ideal for:
 
 > Note: Vanilla mode is a runtime switch and does not change the Agent API or class signatures. It is controlled via CLI/config only.
 
-## 🧑‍💻 AgentProfileManager: Profile, Role, and Prompt Management
+## 👨‍💻 AgentProfileManager: Profile, Role, and Prompt Management
 
-Janito now uses a dedicated `AgentProfileManager` class to manage user profiles, roles, interaction styles, and system prompt selection. This manager:
-- Stores the current role (e.g., "software engineer") and interaction style (e.g., "default", "technical").
-- Renders the system prompt from the appropriate template based on interaction style.
+Janito now uses a dedicated `AgentProfileManager` class to manage user profiles, roles, interaction profiles, and system prompt selection. This manager:
+- Stores the current role (e.g., "software engineer") and interaction profile (e.g., "default", "technical").
+- Renders the system prompt from the appropriate template based on interaction profile.
 - Instantiates and manages the low-level LLM Agent, passing the correct prompt.
-- Provides methods to update the role, interaction style, and refresh the prompt at runtime.
+- Provides methods to update the role, interaction profile, and refresh the prompt at runtime.
 
 ### Multiple System Prompt Templates
-- The system prompt template is now selected based on the interaction style (e.g., `default` or `technical`).
+- The system prompt template is now selected based on the interaction profile (e.g., `default` or `technical`).
 - Templates are located in `janito/agent/templates/` (see `system_prompt_template.j2` and `system_prompt_template_technical.j2`).
-- You can switch interaction styles at runtime using the profile manager, enabling different agent behaviors for different user needs.
+- You can switch interaction profiles at runtime using the profile manager, enabling different agent behaviors for different user needs.
 
 This separation ensures that the LLM Agent remains focused on language model interaction and tool execution, while all profile, role, and prompt logic is managed at a higher level.
 
@@ -174,47 +209,47 @@ See `janito/agent/profile_manager.py` for implementation details.
 
 ### Agent Interaction Style
 
-You can control the agent's behavior and prompt style globally or per-project using the `interaction_style` config key. Supported values:
+You can control the agent's behavior and prompt profile globally or per-project using the `profile` config key. See [Prompt Profiles Guide](docs/guides/prompt_profiles.md) for all available styles and combinations.
 - `default`: Concise, general-purpose agent (default)
 - `technical`: Strict, workflow-oriented for technical/developer use
 
 Set globally:
 ```bash
-janito --set-global-config interaction_style=technical
+janito --set-global-config profile=technical
 ```
 Or per-project (in your project root):
 ```bash
-janito --set-local-config interaction_style=technical
+janito --set-local-config profile=technical
 ```
 
 You can also override for a session with the CLI flag:
 ```bash
-janito --style technical
+janito --profile technical
 ```
 
 See [docs/CONFIGURATION.md](docs/CONFIGURATION.md) for full details.
 
-## 🧩 Combinatorial Style System
+## 🧑‍💻 Combinatorial Style System
 
-Janito now supports combinatorial styles for system prompts, allowing you to combine a main style (such as `default` or `technical`) with one or more feature extensions (such as `commit_all`).
+Janito now supports combinatorial profiles for system prompts, allowing you to combine a main profile (such as `default` or `technical`) with one or more feature extensions (such as `commit_all`).
 
-- **Main style:** The base agent behavior and workflow (e.g., `default`, `technical`).
-- **Feature extensions:** Optional features that override or extend the main style (e.g., `commit_all`).
+- **Main profile:** The base agent behavior and workflow (e.g., `default`, `technical`).
+- **Feature extensions:** Optional features that override or extend the main profile (e.g., `commit_all`).
 - **Syntax:** Use a hyphen to combine, e.g., `technical-commit_all`.
 
 **How it works:**
-- The main style template is loaded first.
+- The main profile template is loaded first.
 - Each feature extension template is layered on top, overriding or extending specific blocks in the main template.
 - Feature templates must use `{% extends parent_template %}` for dynamic inheritance.
 
 **Example usage:**
 ```bash
-janito --style technical-commit_all
+janito --profile technical-commit_all
 ```
 
-This will apply the `technical` style with the `commit_all` feature enabled in the agent's system prompt.
+This will apply the `technical` profile with the `commit_all` feature enabled in the agent's system prompt.
 
 See `janito/render_prompt.py` and `janito/agent/templates/` for implementation details and to create your own feature extensions.
 
 ---
-_generated by janito.dev_
+_generated by janito.dev_

janito-1.9.0/janito/__init__.py

@@ -0,0 +1 @@
+__version__ = "1.9.0"

{janito-1.7.0 → janito-1.9.0}/janito/agent/config.py

@@ -64,7 +64,7 @@ CONFIG_OPTIONS = {
     "use_azure_openai": "Whether to use Azure OpenAI client (default: False)",
     # Accept template.* keys as valid config keys (for CLI validation, etc.)
     "template": "Template context dictionary for Agent Profile prompt rendering (nested)",
-    "interaction_style": "Interaction style for the Agent Profile (e.g., 'default' or 'technical')",
+    "profile": "Agent Profile name (only 'base' is supported)",
     # Note: template.* keys are validated dynamically, not statically here
 }
 

janito-1.9.0/janito/agent/config_defaults.py

@@ -0,0 +1,36 @@
+# Centralized config defaults for Janito
+CONFIG_DEFAULTS = {
+    "api_key": None,  # Must be set by user
+    "model": "openai/gpt-4.1",  # Default model
+    "base_url": "https://openrouter.ai/api/v1",
+    "role": "software developer",  # Part of the Agent Profile
+    "system_prompt_template": None,  # None means auto-generate from Agent Profile role
+    "temperature": 0.2,
+    "max_tokens": 200000,
+    "use_azure_openai": False,
+    "azure_openai_api_version": "2023-05-15",
+    "profile": "base",
+    "providers": {
+        "openai": {
+            "api_key": None,
+            "base_url": "https://api.openai.com/v1",
+            "default_model": "gpt-3.5-turbo",
+        },
+        "azureai": {
+            "api_key": None,
+            "base_url": "https://your-azure-endpoint.openai.azure.com/",
+            "api_version": "2023-05-15",
+            "default_model": "gpt-35-turbo",
+        },
+        "openrouterai": {
+            "api_key": None,
+            "base_url": "https://openrouter.ai/api/v1",
+            "default_model": "openrouter/cognitive",
+        },
+        "fireworksai": {
+            "api_key": None,
+            "base_url": "https://api.fireworks.ai/inference/v1",
+            "default_model": "accounts/fireworks/models/firefunction-v1",
+        },
+    },
+}

janito-1.9.0/janito/agent/conversation.py

@@ -0,0 +1,197 @@
+from janito.agent.conversation_api import (
+    get_openai_response,
+    get_openai_stream_response,
+    retry_api_call,
+)
+from janito.agent.conversation_tool_calls import handle_tool_calls
+from janito.agent.conversation_ui import show_spinner, print_verbose_event
+from janito.agent.conversation_exceptions import (
+    MaxRoundsExceededError,
+    EmptyResponseError,
+    NoToolSupportError,
+)
+from janito.agent.runtime_config import unified_config, runtime_config
+import pprint
+
+
+class ConversationHandler:
+    def __init__(self, client, model):
+        self.client = client
+        self.model = model
+        self.usage_history = []
+
+    @staticmethod
+    def remove_system_prompt(messages):
+        """
+        Return a new messages list with all system prompts removed.
+        """
+        return [msg for msg in messages if msg.get("role") != "system"]
+
+    def handle_conversation(
+        self,
+        messages,
+        max_rounds=100,
+        message_handler=None,
+        verbose_response=False,
+        spinner=False,
+        max_tokens=None,
+        verbose_events=False,
+        stream=False,
+        verbose_stream=False,
+    ):
+        from janito.agent.conversation_history import ConversationHistory
+
+        # Accept either ConversationHistory or a list for backward compatibility
+        if isinstance(messages, ConversationHistory):
+            history = messages
+        else:
+            history = ConversationHistory(messages)
+
+        if len(history) == 0:
+            raise ValueError("No prompt provided in messages")
+
+        resolved_max_tokens = max_tokens
+        if resolved_max_tokens is None:
+            resolved_max_tokens = unified_config.get("max_tokens", 200000)
+        try:
+            resolved_max_tokens = int(resolved_max_tokens)
+        except (TypeError, ValueError):
+            raise ValueError(
+                "max_tokens must be an integer, got: {resolved_max_tokens!r}".format(
+                    resolved_max_tokens=resolved_max_tokens
+                )
+            )
+
+        # If vanilla mode is set and max_tokens was not provided, default to 8000
+        if runtime_config.get("vanilla_mode", False) and max_tokens is None:
+            resolved_max_tokens = 8000
+
+        for _ in range(max_rounds):
+            try:
+                if stream:
+                    # Streaming mode
+                    def get_stream():
+                        return get_openai_stream_response(
+                            self.client,
+                            self.model,
+                            history.get_messages(),
+                            resolved_max_tokens,
+                            verbose_stream=runtime_config.get("verbose_stream", False),
+                            message_handler=message_handler,
+                        )
+
+                    retry_api_call(get_stream)
+                    return None
+                else:
+                    # Non-streaming mode
+                    def api_call():
+                        return get_openai_response(
+                            self.client,
+                            self.model,
+                            history.get_messages(),
+                            resolved_max_tokens,
+                        )
+
+                    if spinner:
+                        response = show_spinner(
+                            "Waiting for AI response...", retry_api_call, api_call
+                        )
+                    else:
+                        response = retry_api_call(api_call)
+            except NoToolSupportError:
+                print(
+                    "⚠️ Endpoint does not support tool use. Proceeding in vanilla mode (tools disabled)."
+                )
+                runtime_config.set("vanilla_mode", True)
+                if max_tokens is None:
+                    runtime_config.set("max_tokens", 8000)
+                    resolved_max_tokens = 8000
+
+                # Remove system prompt for vanilla mode if needed (call this externally when appropriate)
+                # messages = ConversationHandler.remove_system_prompt(messages)
+                # Retry once with tools disabled
+                def api_call_vanilla():
+                    return get_openai_response(
+                        self.client, self.model, messages, resolved_max_tokens
+                    )
+
+                if spinner:
+                    response = show_spinner(
+                        "Waiting for AI response (tools disabled)...",
+                        retry_api_call,
+                        api_call_vanilla,
+                    )
+                else:
+                    response = retry_api_call(api_call_vanilla)
+                    print(
+                        "[DEBUG] OpenAI API raw response (tools disabled):",
+                        repr(response),
+                    )
+            if runtime_config.get("verbose_response", False):
+                pprint.pprint(response)
+            if response is None or not getattr(response, "choices", None):
+                raise EmptyResponseError(
+                    f"No choices in response; possible API or LLM error. Raw response: {response!r}"
+                )
+            choice = response.choices[0]
+            usage = getattr(response, "usage", None)
+            usage_info = (
+                {
+                    # DEBUG: Show usage extraction
+                    "_debug_raw_usage": getattr(response, "usage", None),
+                    "prompt_tokens": getattr(usage, "prompt_tokens", None),
+                    "completion_tokens": getattr(usage, "completion_tokens", None),
+                    "total_tokens": getattr(usage, "total_tokens", None),
+                }
+                if usage
+                else None
+            )
+            event = {"type": "content", "message": choice.message.content}
+            if runtime_config.get("verbose_events", False):
+                print_verbose_event(event)
+            if message_handler is not None and choice.message.content:
+                message_handler.handle_message(event)
+            if not choice.message.tool_calls:
+                agent_idx = len(
+                    [m for m in history.get_messages() if m.get("role") == "agent"]
+                )
+                self.usage_history.append(
+                    {"agent_index": agent_idx, "usage": usage_info}
+                )
+                # Add assistant response to history
+                history.add_message(
+                    {
+                        "role": "assistant",
+                        "content": choice.message.content,
+                    }
+                )
+                return {
+                    "content": choice.message.content,
+                    "usage": usage_info,
+                    "usage_history": self.usage_history,
+                }
+            # Tool calls
+            tool_responses = handle_tool_calls(
+                choice.message.tool_calls, message_handler=message_handler
+            )
+            agent_idx = len(
+                [m for m in history.get_messages() if m.get("role") == "agent"]
+            )
+            self.usage_history.append({"agent_index": agent_idx, "usage": usage_info})
+            # Add assistant response with tool calls
+            history.add_message(
+                {
+                    "role": "assistant",
+                    "content": choice.message.content,
+                    "tool_calls": [tc.to_dict() for tc in choice.message.tool_calls],
+                }
+            )
+            for tool_response in tool_responses:
+                history.add_message(
+                    {
+                        "role": "tool",
+                        "tool_call_id": tool_response["tool_call_id"],
+                        "content": tool_response["content"],
+                    }
+                )
+        raise MaxRoundsExceededError(f"Max conversation rounds exceeded ({max_rounds})")