docs: add man pages for llm_aggregator

- Add `docs/llm_aggregator.1` (user command man page)
  - Document all command-line options (feed aggregation, content
    filtering, LLM API, output formats, TUI interface)
  - Include environment variables, configuration precedence, examples,
    exit status, and dependencies
- Add `docs/llm_aggregator.5` (configuration file man page)
  - Describe TOML config file structure with sections for
    feed_aggregation, filtering, llm_api, system, and output
  - Provide full example configuration and security considerations

feat: add comprehensive test suite, --base-url option, and central defaults package

- Add comprehensive unit test suite across all major packages
  - `internal/cli/args_test.go`: CLI argument tests for all flags
  - `internal/aggregator/aggregator_test.go` and `article_test.go`: feed
    parsing, filtering, sorting
  - `internal/output/formatter_test.go`: JSON, Markdown, and text output
    formatting
  - `internal/processor/processor_test.go`: article processing, keyword
    filtering, sorting
  - `internal/config/config_test.go`: configuration loading and Viper
    integration
  - `internal/defaults/defaults_test.go`: default constants validation
- Add `--base-url` CLI option to configure custom LLM API endpoints
  - Support any OpenAI-compatible provider (e.g., local Ollama, Azure
    OpenAI)
  - Environment variable: `LLM_AGGREGATOR_BASE_URL`
  - Default: `https://api.deepseek.com`
- Add `internal/defaults` package to centralise all default constants
  - `internal/config`, `internal/runtime`, `internal/llm` now use
    `defaults.Default*`
  - Eliminates hard-coded default values scattered across packages
- Add `docs/TESTING.md` with comprehensive testing guide
  - Explains test structure, helper functions, and coverage goals
  - Includes examples of table-driven tests and nil safety testing
- Fix nil pointer safety in `internal/aggregator`
  - Add nil checks for `progressCtx` before each log/warning/debug call
  - `FeedAggregator` now works correctly when progress context is nil
  - Tests explicitly use `NewFeedAggregatorWithProgress` with proper
    context
- Update `README.md` to document `--base-url` flag and environment
  variable
- Update `CHANGELOG.md` with version 0.6.0 changes

feat: add accurate token counting, concurrent feed fetching, and scrollable TUI summary

- Add accurate token counting with tiktoken
  - New `internal/tokeniser` package using
    `github.com/pkoukk/tiktoken-go`
  - BPE tokenisation for precise counting vs rough character estimation
  - Encoding caching and model-aware encoding selection (cl100k_base,
    o200k_base, etc.)
  - `CountTokens()` and `CountMessagesTokens()` functions
  - Fallback to rough estimation when tiktoken initialisation fails
- Add concurrent feed fetching with errgroup
  - Fetch feeds in parallel using `golang.org/x/sync/errgroup`
  - Semaphore limits concurrent requests to 10 to avoid server overload
  - Mutex-protected shared state for thread-safe article collection
  - Partial feed failures do not stop other feeds from being processed
- Add scrollable summary view in TUI
  - Integrate `bubbles/viewport` component for browsing long summaries
  - Keyboard navigation: j/k or arrows (scroll), Space/B (page), g/G
    (start/end)
  - Mouse wheel scrolling support
  - Scroll progress indicator showing position and total lines
- Streamline configuration management
  - Use global Viper instance with automatic precedence handling
  - Simplify configuration flow: Parse → GetViper → BindCLIArgs →
    ViperToRuntime
  - Remove manual precedence logic from `cmd/llm_aggregator.go`
- Replace TUI hex colour codes with ANSI 256-colour palette
- Fix viewport scroll event handling and content pre-wrapping before
  display
- Rename `internal/llm/deepseek.go` to `internal/llm/llm.go` (generalise
  LLM client)
- Update `CHANGELOG.md` (v0.5.0) and `README.md` to document new
  features

refactor: generalise Deepseek-specific naming to generic LLM

- Rename `DeepseekClient` to `LLMClient` in `internal/llm/deepseek.go`
  - Update constructor `NewDeepseekClient` → `NewLLMClient`
  - Change all comments and log messages from "Deepseek" to "LLM"
- Update `internal/runtime/runtime.go` to use `NewLLMClient` instead
- Replace all references to "Deepseek API" with "LLM API" across:
  - `README.md` – section headings, environment variables, examples
  - `configs/config.example.toml` – comments and section titles
  - `internal/cli/args.go` – flag help text, description, environment
    note
  - `internal/config/config.go` – struct comments and bindEnv calls
- Keep default model as `deepseek-chat` and base URL as
  `https://api.deepseek.com`
  - The client remains compatible with Deepseek but clearly supports any
    OpenAI‑compatible LLM provider
- Improve error messages to be provider‑agnostic (e.g., "invalid API
  key" without mentioning Deepseek)

fix: restructure TUI with bubbletea command pattern and progress interface

- Replace manual goroutine and signal handling with native bubbletea
  commands
  - Move runtime execution into `Model.startProcessing()` tea.Cmd
  - Remove `runWithTUI` goroutine orchestration and channel
    synchronisation
- Introduce `progress.Progress` interface and inject into
  `runtime.Runtime`
  - Add `SetStage`, `SetSubStage`, `SetArticleCount` methods to
    `SimpleLogger` and `NoopLogger`
  - Replace internal `logger *progress.Context` with `Progress
    progress.Progress` field
- Simplify `TUIProgress` to only send messages to existing `tea.Program`
  - Remove program creation and runtime execution from `TUIProgress`
  - Use `StageMsg`, `SubStageMsg`, `ArticleCountMsg` for updates
- Update `Model` to hold `*runtime.Runtime` and display spinner
  - Add `spinner` model and tick command for visual feedback
  - Show checkmark on completion, capture elapsed time, and display
    summary inline
- Clean up unused `StepMsg` related functions and redundant blank lines
- Adjust `applyConfiguration` formatting and remove unused imports
  (`os/signal`, `syscall`)

feat: add configuration file and environment variable support

- Add `internal/config` package with Viper-based TOML loading
  - Load configuration from `~/.config/llm_aggregator/config.toml` (XDG
    compliant)
  - Support all aggregation, API, and output options in config file
  - Implement precedence: CLI args > env vars > config file > built‑in
    defaults
- Integrate config loading into `cmd/llm_aggregator.go`
  - Call `config.Load()` before argument parsing
  - Apply config values to runtime with `applyConfiguration()`
  - Gracefully warn if config file exists but fails to load
- Rename environment variable prefix from `DEEPSEEK_API_KEY` to
  `LLM_AGGREGATOR_API_KEY`
  - Update `internal/cli/args.go` help text and environment variable
    reference
  - Update `internal/llm/deepseek.go` to read `LLM_AGGREGATOR_API_KEY`
  - Update `README.md` and `CHANGELOG.md` to reflect new variable name
- Add `configs/config.example.toml` with documented configuration
  options
- Add `internal/config/config_test.go` covering defaults, path
  resolution, save/load, and environment variable precedence
- Update `go.mod` / `go.sum` with new dependencies:
  - `github.com/spf13/viper` for configuration management
  - `github.com/adrg/xdg` for XDG base directory support
  - `github.com/pelletier/go-toml/v2` (indirect via viper)
- Revise `README.md` to document the new configuration system,
  precedence order, and environment variables
- Update `CHANGELOG.md` for version 0.3.0
- Remove unused `internal/config` global flags (`QuietMode`,
  `VerboseMode`) replaced by runtime configuration

feat: add verbose flag to control progress and status messages

- Make CLI silent by default; only final output and errors are displayed
  - Move all progress logs, warnings, and debug messages behind
    `-v/--verbose`
- Introduce logger interface (`progress.Context`) for consistent logging
  - Add `NoopLogger` for silent mode, `SimpleLogger` for verbose output
- Update components to use logger instead of direct `fmt.Print` calls
  - `aggregator.FeedAggregator` accepts logger via
    `NewFeedAggregatorWithProgress`
  - `processor.ContentProcessor` gains `SetLogger` method
  - `llm.DeepseekClient` gains `SetLogger` method
- Modify `runtime.Runtime` to create appropriate logger from `Verbose`
  flag
- Update CLI flag description: `--verbose` now says "Show verbose
  output"
- Remove redundant conditional logging blocks (if verbose) from
  `cmd/llm_aggregator.go`
- Ensure token estimates, filtering summaries, and API call logs only
  appear in verbose mode

feat: initial implementation of LLM RSS aggregator with DeepSeek summarisation

- Add core aggregation engine (`internal/aggregator/`)
  - Parse RSS/Atom/JSON feeds using `gofeed`
  - Extract article metadata (title, link, date, author)
  - Fallback to web scraping with `goquery` for minimal content
- Add content processing and filtering (`internal/processor/`)
  - Keyword-based include/exclude filtering
  - Sort by date, title, or source
  - Token estimation and context preparation
- Add DeepSeek API integration (`internal/llm/`)
  - OpenAI-compatible client for chat completions
  - Configurable model, max tokens, temperature
  - Support custom system prompts
- Add command-line interface (`internal/cli/`)
  - `go-arg` based argument parsing
  - Options for feeds file, prompt, filtering, output formats
  - Environment variable support for API key
- Add output formatting (`internal/output/`)
  - Plain text, GitHub‑flavoured markdown, and JSON outputs
  - Optional inclusion of original articles
- Add terminal user interface (`internal/tui/`)
  - Bubble Tea based progress bar with gradient colours
  - Live article counters and elapsed time
  - Keyboard controls (q/Ctrl+C to quit)
- Add runtime orchestration (`internal/runtime/`)
  - Pipeline execution (aggregate → process → summarise → output)
  - File output support
- Add project documentation
  - `README.md` with usage examples and dependencies
  - `CHANGELOG.md` for version 0.1.0
  - `LICENCE.txt` (EUPL 1.2)
  - Third‑party library READMEs in `docs/` for reference
- Add build configuration
  - `go.mod` with dependencies (gofeed, openai-go, bubbletea, lipgloss,
    go-arg, goquery)
  - `.goreleaser.yaml` for cross‑platform releases
  - `.gitignore` for Go binaries and test artifacts