Fix PDF layout rendering by resetting multi_cell cursor to LMARGIN

.github/ISSUE_TEMPLATE/automated_patches.md

@@ -0,0 +1,22 @@
+---
+name: 'Feature: Automated Patches / Merge Requests'
+about: Template for applying code patches and committing fixes.
+title: 'Feature: Automated Patches / Merge Requests'
+labels: ['help wanted', 'enhancement']
+assignees: ''
+---
+
+## Description
+
+We want to close the loop on codebase remediation by letting the CLI suggest and commit code fixes automatically.
+
+## Goal
+
+Implement a command (or interactive REPL slash command) that takes an AI-generated code patch, applies it to the local codebase, and creates a git branch or commit.
+
+## Requirements
+
+1. Retrieve the code remediation suggested by the LLM.
+2. Write the patch safely to the target file.
+3. Optionally run a validation check (e.g. `pytest` or compiler checks) to verify the patch doesn't break tests.
+4. Automatically create a local feature branch and git commit with the fix.

.github/ISSUE_TEMPLATE/bug_report.md

@@ -0,0 +1,32 @@
+---
+name: Bug Report
+about: Create a report to help us improve.
+title: 'bug: '
+labels: ['bug', 'triage']
+assignees: ''
+---
+
+## Description
+
+A clear and concise description of what the bug is.
+
+## Steps to Reproduce
+
+Steps to reproduce the behavior:
+1. Run CLI command '...'
+2. See error '...'
+
+## Expected Behavior
+
+A clear and concise description of what you expected to happen.
+
+## Environment Details
+
+* OS: [e.g. Ubuntu 22.04, macOS Sonoma]
+* Python Version: [e.g. 3.12.2]
+* Installation Method: [e.g. cli/install.sh, pip install -e cli/]
+* Active LLM Provider: [e.g. Gemini, OpenAI, Claude, Ollama]
+
+## Additional Context
+
+Add any other context or terminal log snippets about the problem here.

.github/ISSUE_TEMPLATE/ci_cd_integration.md

@@ -0,0 +1,21 @@
+---
+name: 'Feature: CI/CD Integration Packages'
+about: Template for building GitHub Actions and GitLab Runner wrappers.
+title: 'Feature: CI/CD Integration Packages (GitHub Actions & GitLab Runner)'
+labels: ['help wanted', 'enhancement']
+assignees: ''
+---
+
+## Description
+
+We want to make it easy for developers to integrate SecureLens into their CI/CD pipelines.
+
+## Goal
+
+Create a reusable GitHub Action and GitLab Runner template that runs the `securelens` CLI inside workflow pipelines.
+
+## Requirements
+
+1. Wrap the CLI command `securelens scan` so it can run inside a container in GitHub Actions.
+2. Implement configuration options to pass the `--ci` and `--fail-on` flags.
+3. Ensure the action outputs logs cleanly and fails the build step with a non-zero exit code if critical or high vulnerabilities are detected.

.github/ISSUE_TEMPLATE/dependency_auditor.md

@@ -0,0 +1,21 @@
+---
+name: 'Feature: Dependency Lockfile Auditor'
+about: Template for scanning package files against the OSV database.
+title: 'Feature: Dependency Lockfile Auditor (securelens audit)'
+labels: ['help wanted', 'enhancement', 'good first issue']
+assignees: ''
+---
+
+## Description
+
+We want to expand the CLI tool's capabilities to scan project dependencies for known vulnerabilities.
+
+## Goal
+
+Add a new CLI command `securelens audit <path>` that scans package descriptors (such as `requirements.txt` or `package.json`) and runs checks against the Open Source Vulnerability (OSV.dev) database API.
+
+## Requirements
+
+1. Parse common package files to extract dependency names and versions.
+2. Submit query requests to the OSV API (`https://api.osv.dev/v1/query`).
+3. Format the results in a clear CLI table using `rich` showing packages, affected versions, and severity.

CONTRIBUTING.md

@@ -6,7 +6,15 @@ Welcome to SecureLens! We appreciate your interest in contributing to the projec
 
 ## Code of Conduct
 
-By participating in this project, you agree to abide by our Code of Conduct. Please read `CODE_OF_CONDUCT.md` to understand the expectations we have for community members.
+By participating in this project, you agree to abide by our Code of Conduct. Please read [CODE_OF_CONDUCT.md](file:///home/Krishna-Singh/securelens-backend/CODE_OF_CONDUCT.md) to understand the expectations we have for community members.
+
+---
+
+## Where to Start?
+
+If you are a first-time contributor or looking for a quick way to get involved, head over to our GitHub Issues tab. We actively maintain labels like:
+* **`good first issue`** - Small, isolated tasks perfect for getting familiar with the repository layout.
+* **`help wanted`** - Features, bugs, or refactoring tasks that are high priority for our current roadmap.
 
 ---
 
@@ -40,22 +48,31 @@ To submit code changes:
 ### 1. Code Style
 
 We enforce standard Python formatting to keep the codebase maintainable and readable:
-* Use standard line length limits (79 or 88 characters).
+* Use standard line length limits (88 characters).
 * Use explicit type hinting for function arguments and return values where possible.
 * Document modules, classes, and public functions with clear docstrings.
 
-### 2. Linting and Formatting
+### 2. Linting, Formatting & Pre-Commit
+
+Run linting checks before committing. We recommend installing our pre-commit hook framework to automate these checks natively from your git environment:
 
-Run linting checks before committing. If you use formatters like `black` or `ruff`:
 ```bash
-# Format the code base
-black app/ cli/ tests/
+# Install the hooks locally
+pre-commit install
 
-# Run static analysis
-ruff check app/ cli/ tests/
+# Manually run checks against all files
+pre-commit run --all-files
 ```
 
-We recommend installing the pre-commit hook framework to automate checks (see `.pre-commit-config.yaml`).
+If you prefer to run individual formatters and linters manually (or via our project shortcuts):
+```bash
+# Run via Makefile automation
+make lint
+
+# Or run manually
+black app/ cli/ tests/
+ruff check app/ cli/ tests/
+```
 
 ### 3. Running the Test Suite
 
@@ -63,6 +80,10 @@ We use `pytest` for unit and integration testing. All tests are located under th
 
 Run the entire test suite locally:
 ```bash
+# Run via Makefile automation
+make test
+
+# Or run manually
 pytest tests/ -v
 ```
 
@@ -73,6 +94,8 @@ pytest tests/test_cli_sync.py -v
 pytest tests/test_cli_pdf.py -v
 ```
 
+Refer to [test_cli_patterns.py](file:///home/Krishna-Singh/securelens-backend/tests/test_cli_patterns.py), [test_cli_sync.py](file:///home/Krishna-Singh/securelens-backend/tests/test_cli_sync.py), and [test_cli_pdf.py](file:///home/Krishna-Singh/securelens-backend/tests/test_cli_pdf.py) for examples of testing our offline scanner, sync clients, and PDF exporters.
+
 Ensure all tests pass and write new tests covering any logic changes or new features you introduce.
 
 ---
@@ -80,7 +103,8 @@ Ensure all tests pass and write new tests covering any logic changes or new feat
 ## Pull Request Guidelines
 
 To help maintainers review your PR efficiently:
-* Keep PRs focused. Do not combine multiple unrelated changes into one pull request.
-* Write a clear title and description explaining what was changed and why.
-* Reference any related issues (e.g., `Closes #123`).
-* Verify that your branch is rebased on top of the latest `main` branch.
+* **Keep PRs focused.** Do not combine multiple unrelated changes into one pull request.
+* **Provide context.** Write a clear title and description explaining what was changed and why.
+* **Link your issues.** Reference any related issues (e.g., `Closes #123`) so they close automatically upon merge.
+* **Stay updated.** Verify that your branch is rebased on top of the latest upstream `main` branch.
+* **CI Validation:** Every PR triggers an automated workflow running our formatting and test suites. PRs cannot be merged until these checks pass.

README.md

@@ -130,18 +130,24 @@ uvicorn app.main:app --reload
 
 ### 2. Install Local CLI
 
-Install the `securelens` CLI globally or inside your workspace virtual environment:
+Choose one of the following installation methods based on your platform and preferences:
+
+#### Method A: One-Line Global Installation (Cross-Platform)
+Recommended for general use. Installs the CLI globally on Windows, macOS, or Linux directly from the repository without needing to clone it manually:
 ```bash
-# Execute the automated installer script
+pip install git+https://github.com/Rarebuffalo/securelens-backend.git#subdirectory=cli
+```
+*(Or use `pipx install git+...` to automatically isolate the tool in its own environment).*
+
+#### Method B: Local Source Installation (Development)
+If you have cloned this repository locally:
+```bash
+# On Linux/macOS
 chmod +x cli/install.sh
 ./cli/install.sh
-
-# Activate environment to use the tool
 source venv/bin/activate
-```
 
-Alternatively, install in editable mode:
-```bash
+# On Windows or manual setup
 pip install -e cli/
 ```
 

SETUP.md

@@ -88,25 +88,26 @@ This starts the containers in the background, maps port 8000 to the host, and pe
 
 The `securelens` CLI allows you to execute scans directly from your local terminal and synchronize reports back to your backend account.
 
-### 1. Automated Script Installation
+Choose one of the following installation methods based on your platform and preferences:
 
-You can install the CLI automatically using the provided installer:
+### Method A: One-Line Global Installation (Cross-Platform)
+Recommended for general use. Installs the CLI globally on Windows, macOS, or Linux directly from the remote repository without needing to clone it:
+```bash
+pip install git+https://github.com/Rarebuffalo/securelens-backend.git#subdirectory=cli
+```
+*(Or use `pipx install git+...` to automatically isolate the tool in its own environment).*
 
+### Method B: Local Source Installation (Development)
+If you are contributing to this codebase or want to install it from the cloned repository folder:
+
+#### 1. Automated Script Installation (Linux/macOS)
 ```bash
 chmod +x cli/install.sh
 ./cli/install.sh
-```
-
-Ensure you have activated your virtual environment before running the tool:
-
-```bash
 source venv/bin/activate
 ```
 
-### 2. Manual Installation
-
-Alternatively, you can install the CLI in editable mode manually:
-
+#### 2. Manual Installation (Windows / Cross-Platform)
 ```bash
 pip install -e cli/
 ```

app/config.py

@@ -55,6 +55,9 @@ class Settings(BaseSettings):
     # Leave blank for Ollama (local, no key needed).
     ai_api_key: str | None = None
 
+    # AI_API_BASE: Custom API base URL (e.g. for Agent Router or custom OpenAI-compatible proxies)
+    ai_api_base: str | None = None
+
     # -------------------------------------------------------------------------
     # Legacy Gemini key — kept for backward compatibility.
     # If AI_API_KEY is not set but GEMINI_API_KEY is, we use that automatically.

app/routers/code_scan.py

@@ -138,7 +138,7 @@ async def chat_with_scan(
     We load the scan from PostgreSQL using the scan_id, so this works
     correctly across server restarts and multiple workers.
     """
-    if not settings.ai_api_key:
+    if not settings.effective_ai_key:
         raise HTTPException(
             status_code=400,
             detail="AI Chat is disabled because no AI API key is configured.",
@@ -285,12 +285,12 @@ async def list_available_models():
     Lists AI models available to the configured provider.
     Only meaningful when using the Gemini provider.
     """
-    if not settings.ai_api_key:
+    if not settings.effective_ai_key:
         raise HTTPException(status_code=500, detail="No AI API key is set.")
     try:
         from google import genai
 
-        client = genai.Client(api_key=settings.ai_api_key)
+        client = genai.Client(api_key=settings.effective_ai_key)
         models = []
         for model in client.models.list():
             if "generateContent" in model.supported_actions:

app/services/ai.py

@@ -74,6 +74,15 @@ async def call_ai(
         "api_key": api_key,
     }
 
+    if settings.ai_api_base:
+        kwargs["api_base"] = settings.ai_api_base
+        if "agentrouter.org" in settings.ai_api_base.lower():
+            kwargs["extra_headers"] = {
+                "Originator": "codex_cli_rs",
+                "User-Agent": "codex_cli_rs/0.101.0 (Mac OS 26.0.1; arm64) Apple_Terminal/464",
+                "Version": "0.101.0",
+            }
+
     # JSON mode: supported natively by OpenAI and LiteLLM proxied Gemini.
     # For providers that don't support it, LiteLLM silently ignores the flag.
     if json_mode:

cli/pyproject.toml

@@ -18,6 +18,7 @@ dependencies = [
     "pyyaml>=6.0",
     "questionary>=2.0",
     "pathspec>=0.12",
+    "fpdf2>=2.7",
 ]
 
 [project.scripts]

cli/securelens/ai/__init__.py

@@ -21,6 +21,7 @@ async def call_ai(
     temperature: float = 0.3,
     json_mode: bool = False,
     conversation_history: Optional[list] = None,
+    api_base: Optional[str] = None,
 ) -> str:
     """
     Single entry-point for all AI calls in the CLI.
@@ -34,6 +35,7 @@ async def call_ai(
     json_mode            : Ask the model to respond with valid JSON only
     conversation_history : Optional list of {"role": ..., "content": ...} dicts
                            for multi-turn chat sessions
+    api_base             : Optional custom API base URL (e.g. for Agent Router)
     """
     import litellm
 
@@ -49,6 +51,15 @@ async def call_ai(
         "api_key": api_key if api_key else None,
     }
 
+    if api_base:
+        kwargs["api_base"] = api_base
+        if "agentrouter.org" in api_base.lower():
+            kwargs["extra_headers"] = {
+                "Originator": "codex_cli_rs",
+                "User-Agent": "codex_cli_rs/0.101.0 (Mac OS 26.0.1; arm64) Apple_Terminal/464",
+                "Version": "0.101.0",
+            }
+
     if json_mode:
         kwargs["response_format"] = {"type": "json_object"}
 
@@ -65,9 +76,10 @@ async def call_ai_json(
     api_key: str,
     model: str,
     temperature: float = 0.2,
+    api_base: Optional[str] = None,
 ) -> Optional[dict]:
     """Convenience wrapper — calls AI in JSON mode and parses the result."""
-    raw = await call_ai(prompt, api_key, model, temperature=temperature, json_mode=True)
+    raw = await call_ai(prompt, api_key, model, temperature=temperature, json_mode=True, api_base=api_base)
     if not raw:
         return None
     try:

cli/securelens/cli.py

@@ -30,8 +30,8 @@ def _run(coro):
 
 
 def _require_config(cfg):
-    """Exit early with a friendly message if no API key is set."""
-    if not cfg.api_key:
+    """Exit early with a friendly message if no API key is set and not using a local model."""
+    if not cfg.api_key and not cfg.default_model.startswith("ollama/"):
         console.print(
             "\n[bold yellow]⚠ No API key configured.[/bold yellow]\n"
             "  Run [bold cyan]securelens configure[/bold cyan] to set one up.\n"
@@ -42,15 +42,17 @@ def _require_config(cfg):
 
 # ── Main group ─────────────────────────────────────────────────────────────────
 
-@click.group()
+@click.group(invoke_without_command=True)
 @click.version_option("2.0.0", prog_name="SecureLens AI")
-def main():
+@click.pass_context
+def main(ctx):
     """
     \b
     SecureLens AI — AI-powered security scanner
     Scan codebases, URLs and get instant security reports.
     """
-    pass
+    if ctx.invoked_subcommand is None:
+        click.echo(ctx.get_help())
 
 
 # ── configure ─────────────────────────────────────────────────────────────────
@@ -74,7 +76,8 @@ def configure():
         "4": ("gpt-4o",                   "OpenAI GPT-4o"),
         "5": ("claude-3-5-haiku-20241022","Anthropic Claude 3.5 Haiku"),
         "6": ("ollama/llama3.1",          "Ollama (local, no key needed)"),
-        "7": ("custom",                   "Custom model string"),
+        "7": ("agentrouter",              "Agent Router (OpenAI-compatible gateway)"),
+        "8": ("custom",                   "Custom LiteLLM model / Endpoint"),
     }
     console.print("[bold]Choose AI Provider:[/bold]")
     for k, (_, desc) in providers.items():
@@ -83,11 +86,21 @@ def configure():
 
     choice = Prompt.ask("Select", choices=list(providers.keys()), default="1")
     model_str, _ = providers[choice]
+    api_base = ""
 
-    if model_str == "custom":
+    if choice == "7":
+        model_str = Prompt.ask("Enter model name (must start with 'openai/')", default="openai/deepseek-chat")
+        if not model_str.startswith("openai/"):
+            model_str = "openai/" + model_str
+        api_base = Prompt.ask("Enter API base URL", default="https://agentrouter.org/v1").strip()
+    elif choice == "8":
+        model_str = Prompt.ask("Enter LiteLLM model string (e.g. openai/my-model-name)")
+        api_base = Prompt.ask("Enter custom API base URL (optional, e.g. https://my-endpoint/v1)", default="").strip()
+    elif model_str == "custom":
         model_str = Prompt.ask("Enter LiteLLM model string (e.g. openrouter/google/gemini-flash)")
 
     cfg.default_model = model_str
+    cfg.api_base = api_base
 
     # API key (skip for Ollama)
     if not model_str.startswith("ollama/"):
@@ -106,6 +119,8 @@ def configure():
     save_config(cfg)
     console.print(f"\n[bold green]✓ Config saved to {CONFIG_FILE}[/bold green]")
     console.print(f"  Model:  [cyan]{cfg.default_model}[/cyan]")
+    if cfg.api_base:
+        console.print(f"  Base URL: [cyan]{cfg.api_base}[/cyan]")
     console.print(f"  Output: [cyan]{cfg.output_format}[/cyan]\n")
 
 
@@ -158,10 +173,32 @@ async def _scan_async(path, model, output, max_files, ci, fail_on, no_ai, sync):
         cfg.max_files_to_scan = max_files
 
     if not no_ai:
-        _require_config(cfg)
+        if not cfg.api_key and not cfg.default_model.startswith("ollama/"):
+            console.print(
+                "\n[bold yellow]⚠ No API key configured.[/bold yellow] Automatically falling back to [bold cyan]offline pattern-based mode[/bold cyan].\n"
+                "  To use AI capabilities, run [bold cyan]securelens configure[/bold cyan] to set an API key,\n"
+                "  or set the [dim]SECURELENS_API_KEY[/dim] environment variable.\n"
+            )
+            no_ai = True
+        else:
+            _require_config(cfg)
 
     root = Path(path).resolve()
 
+    # Safety check: prevent scanning home/root directories by mistake
+    if root == Path.home() or root == Path("/"):
+        if not ci:
+            if not Confirm.ask(
+                f"\n[bold yellow]⚠ Warning: You are attempting to scan your home or root directory ({root}).[/bold yellow]\n"
+                "  This may contain system caches, virtual environments, and massive system files.\n"
+                "  Do you want to continue?"
+            ):
+                console.print("[dim]Scan cancelled.[/dim]\n")
+                sys.exit(0)
+        else:
+            console.print(f"[bold red]✗ Error: Cannot scan home/root directory ({root}) in CI mode.[/bold red]\n")
+            sys.exit(1)
+
     if not ci:
         print_banner()
         print_scan_header(str(root), cfg.default_model)
@@ -232,7 +269,7 @@ async def _scan_async(path, model, output, max_files, ci, fail_on, no_ai, sync):
                 for v in vulnerabilities
             ]
             prompt = summary_prompt(str(root), _json.dumps(issues_data, indent=2))
-            ai_summary = await call_ai(prompt, cfg.api_key, cfg.default_model, temperature=0.4)
+            ai_summary = await call_ai(prompt, cfg.api_key, cfg.default_model, temperature=0.4, api_base=cfg.api_base)
         progress.update(task_summary, completed=100, total=100, detail="Done")
 
     # ── Build result ─────────────────────────────────────────────────────────
@@ -275,13 +312,14 @@ async def _scan_async(path, model, output, max_files, ci, fail_on, no_ai, sync):
         return
 
     # ── Interactive REPL ─────────────────────────────────────────────────────
-    if fmt in ("terminal", "all", "markdown") and not no_ai:
+    if fmt in ("terminal", "all", "markdown"):
         ctx = ReplContext(
             target=str(root),
             scan_result=result,
             target_type="code",
-            api_key=cfg.api_key,
+            api_key=cfg.api_key if not no_ai else None,
             model=cfg.default_model,
+            api_base=cfg.api_base if not no_ai else None,
         )
         await run_repl(ctx)
 
@@ -351,7 +389,7 @@ async def _web_async(url, model, output, ci, fail_on, no_ai):
             ]
             prompt = web_summary_prompt(url, _json.dumps(issues_data, indent=2),
                                         result.score, result.grade)
-            result.ai_summary = await call_ai(prompt, cfg.api_key, cfg.default_model, temperature=0.4)
+            result.ai_summary = await call_ai(prompt, cfg.api_key, cfg.default_model, temperature=0.4, api_base=cfg.api_base)
         progress.update(task2, completed=100, total=100, detail="Done")
 
     fmt = cfg.output_format
@@ -370,13 +408,14 @@ async def _web_async(url, model, output, ci, fail_on, no_ai):
         _ci_exit(result.issues, fail_on, "web")
         return
 
-    if fmt in ("terminal", "all", "markdown") and not no_ai:
+    if fmt in ("terminal", "all", "markdown"):
         ctx = ReplContext(
             target=url,
             scan_result=result,
             target_type="web",
-            api_key=cfg.api_key,
+            api_key=cfg.api_key if not no_ai else None,
             model=cfg.default_model,
+            api_base=cfg.api_base if not no_ai else None,
         )
         await run_repl(ctx)
 

cli/securelens/config.py

@@ -20,6 +20,7 @@ class CLIConfig:
     # AI backend
     default_model: str = "gemini/gemini-2.0-flash"
     api_key: str = ""
+    api_base: str = ""
 
     # Backend Integration (for sync / auth)
     backend_url: str = "http://localhost:8000"
@@ -61,6 +62,7 @@ def load_config() -> CLIConfig:
             data = yaml.safe_load(f) or {}
         cfg.default_model = data.get("default_model", cfg.default_model)
         cfg.api_key = data.get("api_key", cfg.api_key)
+        cfg.api_base = data.get("api_base", cfg.api_base)
         cfg.backend_url = data.get("backend_url", cfg.backend_url)
         cfg.token = data.get("token", cfg.token)
         cfg.output_format = data.get("output_format", cfg.output_format)
@@ -82,6 +84,11 @@ def load_config() -> CLIConfig:
         or os.environ.get("AI_MODEL")
         or cfg.default_model
     )
+    cfg.api_base = (
+        os.environ.get("SECURELENS_API_BASE")
+        or os.environ.get("AI_API_BASE")
+        or cfg.api_base
+    )
 
     return cfg
 
@@ -92,6 +99,7 @@ def save_config(cfg: CLIConfig) -> None:
     data = {
         "default_model": cfg.default_model,
         "api_key": cfg.api_key,
+        "api_base": cfg.api_base,
         "backend_url": cfg.backend_url,
         "token": cfg.token,
         "output_format": cfg.output_format,

cli/securelens/output/pdf.py

@@ -2,6 +2,39 @@ from fpdf import FPDF
 import datetime
 from typing import Optional
 
+def sanitize_text(text: Optional[str]) -> str:
+    if not text:
+        return ""
+    replacements = {
+        "\u2018": "'",
+        "\u2019": "'",
+        "\u201c": '"',
+        "\u201d": '"',
+        "\u2013": "-",
+        "\u2014": "-",
+        "\u2022": "*",
+        "\u2026": "...",
+        "\u2713": "OK",
+        "\u2714": "OK",
+        "\u2715": "X",
+        "\u2717": "X",
+        "\u2718": "X",
+        "\u26a0": "!",
+        "\u25b6": ">",
+        "\u25c0": "<",
+        "\u25b2": "^",
+        "\u25bc": "v",
+        "\u25ae": "|",
+        "\u2588": "#",
+        "\u2591": ".",
+        "\u2592": ":",
+        "\u2593": "#",
+        "`": "'",
+    }
+    for orig, rep in replacements.items():
+        text = text.replace(orig, rep)
+    return text.encode("latin-1", errors="replace").decode("latin-1")
+
 class SecureLensPDF(FPDF):
     def footer(self):
         self.set_y(-15)
@@ -26,9 +59,9 @@ def export_code_pdf(result, output_path: str) -> str:
     pdf.set_font("helvetica", "", 10)
     pdf.set_text_color(100, 100, 100)
     now_str = datetime.datetime.now().strftime("%Y-%m-%d %H:%M:%S")
-    pdf.cell(0, 8, f"Target Path: {result.target}", new_x="LMARGIN", new_y="NEXT")
+    pdf.cell(0, 8, sanitize_text(f"Target Path: {result.target}"), new_x="LMARGIN", new_y="NEXT")
     pdf.cell(0, 8, f"Scan Time: {now_str}", new_x="LMARGIN", new_y="NEXT")
-    pdf.cell(0, 8, f"Security Score: {result.score}/100 (Grade: {result.grade})", new_x="LMARGIN", new_y="NEXT")
+    pdf.cell(0, 8, sanitize_text(f"Security Score: {result.score}/100 (Grade: {result.grade})"), new_x="LMARGIN", new_y="NEXT")
     pdf.ln(5)
 
     # Executive Summary Section
@@ -40,7 +73,7 @@ def export_code_pdf(result, output_path: str) -> str:
     pdf.set_text_color(0, 0, 0)
     pdf.ln(2)
     summary_text = result.ai_summary or f"A static patterns analysis was performed on the codebase. Out of the files discovered, {len(result.vulnerabilities)} potential security vulnerabilities were reported."
-    pdf.multi_cell(0, 5, summary_text)
+    pdf.multi_cell(0, 5, sanitize_text(summary_text), new_x="LMARGIN", new_y="NEXT")
     pdf.ln(8)
 
     # Files Scanned Section
@@ -53,7 +86,7 @@ def export_code_pdf(result, output_path: str) -> str:
     files_list = ", ".join(result.files_triaged[:15])
     if len(result.files_triaged) > 15:
         files_list += f", and {len(result.files_triaged) - 15} more"
-    pdf.multi_cell(0, 5, files_list or "No files selected.")
+    pdf.multi_cell(0, 5, sanitize_text(files_list or "No files selected."), new_x="LMARGIN", new_y="NEXT")
     pdf.ln(8)
 
     # Issues Findings Section
@@ -78,26 +111,26 @@ def export_code_pdf(result, output_path: str) -> str:
             else: pdf.set_text_color(0, 100, 0)
             
             line_str = f" [Line {v.line_number}]" if v.line_number else ""
-            pdf.cell(0, 8, f"{idx}. {severity}: {v.issue}{line_str}", new_x="LMARGIN", new_y="NEXT")
+            pdf.cell(0, 8, sanitize_text(f"{idx}. {severity}: {v.issue}{line_str}"), new_x="LMARGIN", new_y="NEXT")
             
             # Details
             pdf.set_text_color(0, 0, 0)
             pdf.set_font("helvetica", "B", 9)
             pdf.cell(20, 6, "File:", border=0)
             pdf.set_font("helvetica", "", 9)
-            pdf.cell(0, 6, v.file_path, new_x="LMARGIN", new_y="NEXT")
+            pdf.cell(0, 6, sanitize_text(v.file_path), new_x="LMARGIN", new_y="NEXT")
             
             pdf.set_font("helvetica", "B", 9)
             pdf.cell(0, 6, "Explanation:", new_x="LMARGIN", new_y="NEXT")
             pdf.set_font("helvetica", "", 9)
-            pdf.multi_cell(0, 4.5, v.explanation)
+            pdf.multi_cell(0, 4.5, sanitize_text(v.explanation), new_x="LMARGIN", new_y="NEXT")
             
             if v.suggested_fix:
                 pdf.set_font("helvetica", "B", 9)
                 pdf.cell(0, 6, "Suggested Fix:", new_x="LMARGIN", new_y="NEXT")
                 pdf.set_font("courier", "", 8.5)
                 pdf.set_fill_color(245, 245, 245)
-                pdf.multi_cell(0, 4.5, v.suggested_fix, fill=True)
+                pdf.multi_cell(0, 4.5, sanitize_text(v.suggested_fix), fill=True, new_x="LMARGIN", new_y="NEXT")
             
             pdf.ln(4)
             pdf.line(pdf.get_x(), pdf.get_y(), 200, pdf.get_y())
@@ -123,9 +156,9 @@ def export_web_pdf(result, output_path: str) -> str:
     pdf.set_font("helvetica", "", 10)
     pdf.set_text_color(100, 100, 100)
     now_str = datetime.datetime.now().strftime("%Y-%m-%d %H:%M:%S")
-    pdf.cell(0, 8, f"Target URL: {result.url}", new_x="LMARGIN", new_y="NEXT")
+    pdf.cell(0, 8, sanitize_text(f"Target URL: {result.url}"), new_x="LMARGIN", new_y="NEXT")
     pdf.cell(0, 8, f"Scan Time: {now_str}", new_x="LMARGIN", new_y="NEXT")
-    pdf.cell(0, 8, f"Security Score: {result.score}/100 (Grade: {result.grade})", new_x="LMARGIN", new_y="NEXT")
+    pdf.cell(0, 8, sanitize_text(f"Security Score: {result.score}/100 (Grade: {result.grade})"), new_x="LMARGIN", new_y="NEXT")
     if result.ssl_expiry_days is not None:
         pdf.cell(0, 8, f"SSL Expiry: {result.ssl_expiry_days} days left", new_x="LMARGIN", new_y="NEXT")
     pdf.ln(5)
@@ -139,7 +172,7 @@ def export_web_pdf(result, output_path: str) -> str:
     pdf.set_text_color(0, 0, 0)
     pdf.ln(2)
     summary_text = result.ai_summary or f"An automated live security audit was performed on {result.url}. Out of the layers checked, {len(result.issues)} potential issues were flagged."
-    pdf.multi_cell(0, 5, summary_text)
+    pdf.multi_cell(0, 5, sanitize_text(summary_text), new_x="LMARGIN", new_y="NEXT")
     pdf.ln(8)
 
     # Issues Section
@@ -162,20 +195,20 @@ def export_web_pdf(result, output_path: str) -> str:
             elif severity == "Warning": pdf.set_text_color(218, 165, 32)
             else: pdf.set_text_color(0, 100, 0)
             
-            pdf.cell(0, 8, f"{idx}. {severity}: {i.issue}", new_x="LMARGIN", new_y="NEXT")
+            pdf.cell(0, 8, sanitize_text(f"{idx}. {severity}: {i.issue}"), new_x="LMARGIN", new_y="NEXT")
             
             # Details
             pdf.set_text_color(0, 0, 0)
             pdf.set_font("helvetica", "B", 9)
             pdf.cell(20, 6, "Layer:", border=0)
             pdf.set_font("helvetica", "", 9)
-            pdf.cell(0, 6, i.layer, new_x="LMARGIN", new_y="NEXT")
+            pdf.cell(0, 6, sanitize_text(i.layer), new_x="LMARGIN", new_y="NEXT")
             
             pdf.set_font("helvetica", "B", 9)
             pdf.cell(0, 6, "Remediation / Fix:", new_x="LMARGIN", new_y="NEXT")
             pdf.set_font("courier", "", 8.5)
             pdf.set_fill_color(245, 245, 245)
-            pdf.multi_cell(0, 4.5, i.fix, fill=True)
+            pdf.multi_cell(0, 4.5, sanitize_text(i.fix), fill=True, new_x="LMARGIN", new_y="NEXT")
             
             pdf.ln(4)
             pdf.line(pdf.get_x(), pdf.get_y(), 200, pdf.get_y())

cli/securelens/repl.py

@@ -57,6 +57,7 @@ class ReplContext:
     target_type: str           # "code" | "web" | "github"
     api_key: str
     model: str
+    api_base: Optional[str] = None
     conversation_history: list = field(default_factory=list)
 
 
@@ -93,7 +94,7 @@ async def run_repl(ctx: ReplContext) -> None:
             continue
 
         # ── AI response ─────────────────────────────────────────────────────
-        if not ctx.api_key:
+        if not ctx.api_key and not ctx.model.startswith("ollama/"):
             console.print(
                 "\n  [bold red]✗ No API key configured.[/bold red] "
                 "Run [cyan]securelens configure[/cyan] to set one.\n"
@@ -109,6 +110,7 @@ async def run_repl(ctx: ReplContext) -> None:
                 model=ctx.model,
                 temperature=0.5,
                 conversation_history=ctx.conversation_history,
+                api_base=ctx.api_base,
             )
 
         if response:

cli/securelens/scanners/__init__.py

@@ -99,6 +99,7 @@ def discover_files(root: Path, cfg: CLIConfig) -> list[Path]:
     Respects .gitignore in the root and cfg.ignore_patterns.
     Skips binaries and files larger than cfg.max_file_size_kb.
     """
+    import os
     # Build a combined spec from config ignore_patterns + .gitignore
     ignore_patterns = list(cfg.ignore_patterns)
     gitignore_path = root / ".gitignore"
@@ -113,18 +114,47 @@ def discover_files(root: Path, cfg: CLIConfig) -> list[Path]:
     spec = pathspec.PathSpec.from_lines("gitwildmatch", ignore_patterns)
     max_bytes = cfg.max_file_size_kb * 1024
 
+    # Hardcoded directory blacklist to prune execution paths immediately
+    prune_dirs = {
+        ".git", "node_modules", "venv", ".venv", "__pycache__",
+        "dist", "build", ".next", ".cache", ".npm", ".cargo",
+        ".rustup", ".local", ".ssh", ".gnupg", ".docker", ".vscode",
+        ".idea", "Library", "Pictures", "Music", "Videos", "Documents"
+    }
+
     candidates: list[Path] = []
-    for p in root.rglob("*"):
-        if not p.is_file():
-            continue
-        rel = p.relative_to(root).as_posix()
-        if spec.match_file(rel):
-            continue
-        if p.suffix.lower() in BINARY_EXTENSIONS:
-            continue
-        if p.stat().st_size > max_bytes:
-            continue
-        candidates.append(p)
+    for dirpath, dirnames, filenames in os.walk(root):
+        # 1. Prune standard blacklisted folders in-place
+        dirnames[:] = [d for d in dirnames if d not in prune_dirs]
+
+        # 2. Prune directories matching the ignore spec
+        active_dirs = []
+        for d in dirnames:
+            rel_path = os.path.relpath(os.path.join(dirpath, d), root)
+            if not spec.match_file(rel_path + "/"):
+                active_dirs.append(d)
+        dirnames[:] = active_dirs
+
+        # 3. Process files in the active directory
+        for f in filenames:
+            p = Path(dirpath) / f
+            rel = p.relative_to(root).as_posix()
+            if spec.match_file(rel):
+                continue
+            if p.suffix.lower() in BINARY_EXTENSIONS:
+                continue
+            try:
+                if p.stat().st_size > max_bytes:
+                    continue
+            except OSError:
+                continue
+            candidates.append(p)
+
+            # Capping safeguard: limit to 1000 candidate files
+            if len(candidates) >= 1000:
+                break
+        if len(candidates) >= 1000:
+            break
 
     return sorted(candidates)
 
@@ -165,7 +195,7 @@ async def triage_files(
     if rel_paths and remaining_budget > 0 and cfg.api_key:
         file_list_str = "\n".join(rel_paths[:300])  # cap to ~300 paths for token budget
         prompt = triage_prompt(file_list_str, remaining_budget)
-        result = await call_ai_json(prompt, cfg.api_key, cfg.default_model, temperature=0.1)
+        result = await call_ai_json(prompt, cfg.api_key, cfg.default_model, temperature=0.1, api_base=cfg.api_base)
         if result and "critical_files" in result:
             for rel in result["critical_files"]:
                 abs_path = root / rel
@@ -203,7 +233,7 @@ async def analyze_file(
         content = content[:30_000] + "\n... (truncated)"
 
     prompt = analysis_prompt(rel, content)
-    result = await call_ai_json(prompt, cfg.api_key, cfg.default_model, temperature=0.2)
+    result = await call_ai_json(prompt, cfg.api_key, cfg.default_model, temperature=0.2, api_base=cfg.api_base)
     if not result:
         return []
 

tests/test_cli_api_base.py

@@ -0,0 +1,118 @@
+from unittest.mock import AsyncMock, patch
+import pytest
+from securelens.ai import call_ai
+
+@pytest.fixture(autouse=True)
+def setup_db():
+    pass
+
+@pytest.mark.asyncio
+async def test_call_ai_passes_api_base():
+    with patch("litellm.acompletion", new_callable=AsyncMock) as mock_acompletion:
+        mock_acompletion.return_value.choices = [
+            AsyncMock(message=AsyncMock(content="Mock response"))
+        ]
+        
+        await call_ai(
+            prompt="Hello",
+            api_key="mock_key",
+            model="openai/deepseek-chat",
+            api_base="https://agentrouter.org/v1"
+        )
+        
+        mock_acompletion.assert_called_once()
+        called_kwargs = mock_acompletion.call_args[1]
+        assert called_kwargs["api_base"] == "https://agentrouter.org/v1"
+        assert called_kwargs["model"] == "openai/deepseek-chat"
+        assert called_kwargs["api_key"] == "mock_key"
+
+@pytest.mark.asyncio
+async def test_backend_call_ai_passes_api_base():
+    from app.services.ai import call_ai as backend_call_ai
+    from app.config import settings
+    
+    with patch("litellm.acompletion", new_callable=AsyncMock) as mock_acompletion:
+        mock_acompletion.return_value.choices = [
+            AsyncMock(message=AsyncMock(content="Mock response"))
+        ]
+        
+        original_key = settings.ai_api_key
+        original_base = settings.ai_api_base
+        
+        try:
+            settings.ai_api_key = "mock_key"
+            settings.ai_api_base = "https://agentrouter.org/v1"
+            
+            await backend_call_ai(prompt="Hello")
+            
+            mock_acompletion.assert_called_once()
+            called_kwargs = mock_acompletion.call_args[1]
+            assert called_kwargs["api_base"] == "https://agentrouter.org/v1"
+            assert called_kwargs["api_key"] == "mock_key"
+        finally:
+            settings.ai_api_key = original_key
+            settings.ai_api_base = original_base
+
+@pytest.mark.asyncio
+async def test_triage_files_passes_api_base():
+    from securelens.scanners import triage_files
+    from securelens.config import CLIConfig
+    from pathlib import Path
+    
+    cfg = CLIConfig()
+    cfg.api_key = "mock_key"
+    cfg.api_base = "https://agentrouter.org/v1"
+    cfg.default_model = "openai/deepseek-chat"
+    
+    with patch("securelens.scanners.call_ai_json", new_callable=AsyncMock) as mock_call_ai_json:
+        mock_call_ai_json.return_value = {"critical_files": []}
+        
+        await triage_files([Path("test.py")], Path("."), cfg)
+        
+        mock_call_ai_json.assert_called_once()
+        called_kwargs = mock_call_ai_json.call_args[1]
+        assert called_kwargs["api_base"] == "https://agentrouter.org/v1"
+
+@pytest.mark.asyncio
+async def test_analyze_file_passes_api_base():
+    from securelens.scanners import analyze_file
+    from securelens.config import CLIConfig
+    from pathlib import Path
+    
+    cfg = CLIConfig()
+    cfg.api_key = "mock_key"
+    cfg.api_base = "https://agentrouter.org/v1"
+    cfg.default_model = "openai/deepseek-chat"
+    
+    mock_file = Path("test.py")
+    with patch("pathlib.Path.read_text", return_value="print('hello')"):
+        with patch("securelens.scanners.call_ai_json", new_callable=AsyncMock) as mock_call_ai_json:
+            mock_call_ai_json.return_value = {"vulnerabilities": []}
+            
+            await analyze_file(mock_file, Path("."), cfg)
+            
+            mock_call_ai_json.assert_called_once()
+            called_kwargs = mock_call_ai_json.call_args[1]
+            assert called_kwargs["api_base"] == "https://agentrouter.org/v1"
+
+@pytest.mark.asyncio
+async def test_call_ai_injects_agentrouter_headers():
+    with patch("litellm.acompletion", new_callable=AsyncMock) as mock_acompletion:
+        mock_acompletion.return_value.choices = [
+            AsyncMock(message=AsyncMock(content="Mock response"))
+        ]
+        
+        await call_ai(
+            prompt="Hello",
+            api_key="mock_key",
+            model="openai/deepseek-chat",
+            api_base="https://agentrouter.org/v1"
+        )
+        
+        mock_acompletion.assert_called_once()
+        called_kwargs = mock_acompletion.call_args[1]
+        assert called_kwargs["extra_headers"] == {
+            "Originator": "codex_cli_rs",
+            "User-Agent": "codex_cli_rs/0.101.0 (Mac OS 26.0.1; arm64) Apple_Terminal/464",
+            "Version": "0.101.0",
+        }

tests/test_cli_pdf.py

@@ -14,15 +14,15 @@ def test_export_code_pdf_compiles(tmp_path):
         VulnerabilityFinding(
             file_path="app.py",
             severity="Critical",
-            issue="Hardcoded Secret Key",
-            explanation="Exposing secret key inside app.py.",
-            suggested_fix="Load key from environment",
+            issue="Hardcoded Secret Key with unicode ’smart’ quotes",
+            explanation="Exposing secret key inside app.py • vulnerable to attacks.",
+            suggested_fix="Load key from environment: jwt_secret = Field(default=\"\") \u25b6 check it.",
             line_number=5
         ),
         VulnerabilityFinding(
             file_path="db.py",
             severity="High",
-            issue="Raw SQL Statement",
+            issue="Raw SQL Statement \u2717 check fail",
             explanation="SQL injection inside db.py.",
             suggested_fix="Use parameterized queries",
             line_number=20
@@ -34,7 +34,7 @@ def test_export_code_pdf_compiles(tmp_path):
         total_files_found=10,
         files_triaged=["app.py", "db.py"],
         vulnerabilities=findings,
-        ai_summary="This is a dummy AI report summary describing security posture."
+        ai_summary="This is a dummy AI report summary describing security posture with check \u2713 and block \u2588."
     )
     result.compute_score()