1# Changelog
  2
  3All notable changes to this project will be documented in this file.
  4
  5The format is based on [Keep a
  6Changelog](https://keepachangelog.com/en/1.1.0/), and this project adheres to
  7[Semantic Versioning](https://semver.org/spec/v2.0.0.html).
  8
  9## [0.7.1] - 2026-04-22
 10
 11### Fixed
 12
 13- Environment variables were correctly overriding config file values, but CLI
 14  arguments without explicit flags were also overriding config file values
 15  unintentionally.
 16  - Root cause: `go-arg` was populating struct fields with default values
 17    from `default:` tags even when those CLI flags weren't provided
 18  - Fix: Removed `default:` tags from Args struct fields for LLM options
 19    (`model`, `base-url`, `api-key`, `max-tokens`, `temperature`);
 20    Viper's own default handling is sufficient
 21  - Fix: Changed Args struct fields from value types to pointer types
 22    (`*string`, `*int`, `*float64`) so that "not provided on CLI" can be
 23    distinguished from "provided but zero/empty"
 24  - Fix: Updated `BindCLIArgs()` and `isZero()` to properly handle pointer
 25    types and distinguish nil pointers from zero values
 26- `isZero()` with compound type switch cases: Compound type cases like `case
 27  *int, *int8, *int16, *int32, *int64` don't properly handle nil comparison when
 28  stored in an interface variable. Fixed by splitting into individual type cases.
 29- `isZero()` missing nil interface handling: When a nil pointer is stored
 30  in an `any` interface, `value != nil` returns `true` (the interface itself
 31  is not nil). Added early `if v == nil { return true }` check.
 32
 33## [0.7.0] - 2026-04-22
 34
 35### Added
 36
 37- **`--dry-run` option**: Validate configuration and preview what would happen
 38  without making any LLM API calls. Useful for validating feeds file and
 39  configuration, checking article counts and filtering results and anything
 40  else that can be done "offline".
 41- **Styled terminal output using `lipgloss`**: All CLI output now uses ANSI
 42  colour codes for better readability:
 43- **`$NO_COLOR` environment variable support**: Respects the ["no
 44color"](https://no-color.org/) standard. When `NO_COLOR` is set, all styling is
 45disabled and plain text is output instead.
 46
 47### Changed
 48
 49- API key validation now only occurs when not using `--dry-run` mode
 50
 51## [0.6.0] - 2026-04-22
 52
 53### Added
 54
 55- **Comprehensive unit test suite**, including CLI argument tests
 56  (`--feeds-file`, `--prompt`, `--api-key`, `--model`, `--base-url`),
 57  configuration tests, aggregator tests, output formatter tests, processor
 58  tests and defaults tests.
 59- **`--base-url` CLI option**: configure custom API endpoints for different LLM
 60  providers. Should support any OpenAI-compatible provider.
 61  - Environment variable: `LLM_AGGREGATOR_BASE_URL`
 62- `internal/defaults` package
 63- `docs/TESTING.md` which contains a comprehensive testing guide with test
 64  descriptions
 65
 66### Changed
 67
 68- **Configuration defaults**: All packages now use `defaults.Default*` constants
 69  - `internal/config`: Uses `defaults.Default*` for default values
 70  - `internal/runtime`: Uses `defaults.Default*` for runtime defaults
 71  - `internal/llm`: Uses `defaults.Default*` for LLM client defaults
 72
 73### Fixed
 74
 75- `nil` pointer safety in aggregator
 76  - Added nil checks for `progressCtx` across all log statements
 77  - `FeedAggregator` now works correctly when progress context is nil
 78  - Tests use `NewFeedAggregatorWithProgress` with explicit progress context
 79
 80## [0.5.0] - 2026-04-22
 81
 82### Added
 83
 84- **Accurate token counting with tiktoken**
 85  - New `internal/tokeniser` package using OpenAI's tiktoken-go library
 86  - BPE tokenisation for accurate token counting vs rough character estimation
 87  - Encoding caching to avoid repeated initialisation overhead
 88  - Model-aware encoding selection (cl100k_base, o200k_base, etc.)
 89  - `CountTokens()` and `CountMessagesTokens()` functions
 90  - Fallback to rough estimation when tiktoken initialisation fails
 91- **Concurrent feed fetching**: Feeds are now fetched concurrently using
 92  `golang.org/x/sync/errgroup`
 93  - Semaphore limits concurrent requests to 10 to avoid server overload
 94  - Mutex-protected shared state for thread-safe article collection
 95  - Partial failures don't stop other feeds from being processed
 96- **Scrollable summary in TUI**
 97  - Bubble Tea viewport component for browsing long summaries
 98  - Keyboard navigation: j/k or arrows (scroll), Space/B (page), g/G
 99  (start/end)
100  - Mouse wheel scrolling support
101  - Scroll progress indicator showing position and total lines
102
103### Changed
104
105- **Streamlined configuration management**: Viper now uses global instance with
106automatic precedence handling. Simplified configuration flow: Parse → GetViper
107→ BindCLIArgs → ViperToRuntime
108- **TUI colour scheme**: All hex colour codes replaced with ANSI 256-colour
109  palette
110
111### Fixed
112
113- Viewport now properly handles scroll events via `Update()`. Summary content
114  pre-wrapped to viewport width before display
115- Fixed string literal colour names to use actual variables
116
117## [0.4.0] - 2026-04-21
118
119### Changed 
120
121- **TUI is now fully functional and production-ready**: Replaced the
122work-in-progress TUI implementation with a robust Bubble Tea command pattern:
123  - Runtime execution now uses native `tea.Cmd` instead of manual goroutine
124  orchestration.
125  - Added real-time progress updates via `progress.Progress` interface with
126  stage, substage, and article count messages.
127  - TUI displays a spinner, elapsed time, and final summary inline.
128  - Removed signal handling and channel synchronisation in favour of Bubble
129  Tea's built-in lifecycle.
130  - `TUIProgress` now only sends messages to an existing program, simplifying
131  integration.
132  - Both verbose mode (`SimpleLogger`) and TUI mode share the same
133  `progress.Progress` interface.
134
135## [0.3.0] - 2026-04-21
136
137### Added
138
139- **Configuration file support via TOML** (XDG-compliant and cross-platform)
140  - Load settings from `~/.config/llm_aggregator/config.toml`
141  - Supports all aggregation, API, and output options
142  - Example configuration file at `configs/config.example.toml`
143- **Environment variable support with `LLM_AGGREGATOR_` prefix**
144  - All configuration options can be set via environment variables
145  - Precedence order: CLI arguments > environment variables > config file >
146  built‑in defaults
147- New `internal/config` package with Viper integration
148  - `Load()` and `Save()` methods for configuration management
149  - `GetConfigPath()` for XDG‑compliant config file location
150  - `ConfigExists()` to check for existing configuration
151- Comprehensive unit tests for configuration loading, saving, and environment
152  variable precedence (`internal/config/config_test.go`)
153
154### Changed
155
156- **Environment variable renamed:** `DEEPSEEK_API_KEY` →
157  `LLM_AGGREGATOR_API_KEY`
158- **CLI help text** now reflects the new API key environment variable and
159  general configuration options
160- **Command‑line argument precedence** implemented in `cmd/llm_aggregator.go`
161via `applyConfiguration()`
162- Updated all help text, documentation, and code references
163- `README.md` completely revised with a new "Configuration" section
164
165### Removed
166
167- Global `QuietMode` and `VerboseMode` variables from
168`internal/config/config.go`. Replaced by runtime‑specific configuration in
169`runtime.Runtime` and CLI arguments
170- Hardcoded system prompt in `runtime.NewRuntime()`. Now left empty, allowing
171configuration or client default to be used
172- CLI flags now correctly override config file and environment variables
173
174## [0.2.0] - 2026-04-21
175
176### Changed
177
178- Moved all progress and status messages behind `-v/--verbose` flag
179- Default CLI mode is now silent except for final output and errors
180- Components (aggregator, processor, LLM client) use logger interface
181controlled by verbose flag
182
183## [0.1.0] - 2026-04-21
184
185### Added
186
187- Complete translation from Python prototype to Go implementation
188- RSS/Atom feed parsing using `gofeed` library
189- DeepSeek API integration via `openai-go` client, configured for `/chat/completions` endpoint
190- Content filtering and processing with keyword‑based include/exclude, date filtering, and sorting
191- Command‑line interface using `go‑arg` with automatic help and version flags
192- Multiple output formats: plain text, GitHub‑flavoured markdown, and JSON
193- Terminal user interface (TUI) built with `bubbletea` featuring:
194  - Animated progress bar with gradient colours (`#FF7CCB` to `#FDFF8C`)
195  - Live article counters (aggregated/processed)
196  - Elapsed time display
197  - Coloured status indicators using `lipgloss` styling
198  - Keyboard controls (q/Ctrl+C to quit)
199- Web content extraction fallback using `goquery` when feed descriptions are minimal
200- Configurable limits: articles per feed, maximum age, total articles
201- Token estimation and API usage logging
202- Environment variable support (`LLM_AGGREGATOR_API_KEY`) for authentication
203- Example feeds file with technology, programming, and free software sources
204
205### Changed
206
207- Program structure organised into standard Go layout: `cmd/`, `internal/`, `pkg/`
208- English spelling conventions maintained throughout (colour, initialise, summarise)
209- Error handling improved with specific messages for common API failures (invalid key, rate limits, etc.)
210- Progress reporting unified through a `ProgressContext` interface for both TUI and CLI modes
211
212### Fixed
213
214- Initial API endpoint mismatch (using `/responses` instead of `/chat/completions`)
215- Nil pointer dereferences when handling optional feed metadata (author, dates)
216- String repetition syntax errors in Go (replaced `"="*80` with `strings.Repeat`)
217- CLI help/version flag handling to show information without requiring other arguments
218- Type compatibility issues with `openai-go` v3 API (message parameters, token usage fields)