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## [0.8.0] - 2026-04-23
 10
 11### Added
 12
 13- **Thinking tags toggle in TUI**: LLMs that output `<think>`/`</think>`
 14  thinking tags now have those sections automatically removed from the summary
 15  display by default. Users can press `t` to toggle the thinking section back
 16  on, displayed in gray at the head of the output with XML tags stripped for
 17  cleaner presentation. A status indicator shows whether thinking is currently
 18  visible.
 19
 20## [0.7.1] - 2026-04-22
 21
 22### Fixed
 23
 24- Environment variables were correctly overriding config file values, but CLI
 25  arguments without explicit flags were also overriding config file values
 26  unintentionally.
 27  - Root cause: `go-arg` was populating struct fields with default values
 28    from `default:` tags even when those CLI flags weren't provided
 29  - Fix: Removed `default:` tags from Args struct fields for LLM options
 30    (`model`, `base-url`, `api-key`, `max-tokens`, `temperature`);
 31    Viper's own default handling is sufficient
 32  - Fix: Changed Args struct fields from value types to pointer types
 33    (`*string`, `*int`, `*float64`) so that "not provided on CLI" can be
 34    distinguished from "provided but zero/empty"
 35  - Fix: Updated `BindCLIArgs()` and `isZero()` to properly handle pointer
 36    types and distinguish nil pointers from zero values
 37- `isZero()` with compound type switch cases: Compound type cases like `case
 38  *int, *int8, *int16, *int32, *int64` don't properly handle nil comparison when
 39  stored in an interface variable. Fixed by splitting into individual type cases.
 40- `isZero()` missing nil interface handling: When a nil pointer is stored
 41  in an `any` interface, `value != nil` returns `true` (the interface itself
 42  is not nil). Added early `if v == nil { return true }` check.
 43
 44## [0.7.0] - 2026-04-22
 45
 46### Added
 47
 48- **`--dry-run` option**: Validate configuration and preview what would happen
 49  without making any LLM API calls. Useful for validating feeds file and
 50  configuration, checking article counts and filtering results and anything
 51  else that can be done "offline".
 52- **Styled terminal output using `lipgloss`**: All CLI output now uses ANSI
 53  colour codes for better readability:
 54- **`$NO_COLOR` environment variable support**: Respects the ["no
 55color"](https://no-color.org/) standard. When `NO_COLOR` is set, all styling is
 56disabled and plain text is output instead.
 57
 58### Changed
 59
 60- API key validation now only occurs when not using `--dry-run` mode
 61
 62## [0.6.0] - 2026-04-22
 63
 64### Added
 65
 66- **Comprehensive unit test suite**, including CLI argument tests
 67  (`--feeds-file`, `--prompt`, `--api-key`, `--model`, `--base-url`),
 68  configuration tests, aggregator tests, output formatter tests, processor
 69  tests and defaults tests.
 70- **`--base-url` CLI option**: configure custom API endpoints for different LLM
 71  providers. Should support any OpenAI-compatible provider.
 72  - Environment variable: `LLM_AGGREGATOR_BASE_URL`
 73- `internal/defaults` package
 74- `docs/TESTING.md` which contains a comprehensive testing guide with test
 75  descriptions
 76
 77### Changed
 78
 79- **Configuration defaults**: All packages now use `defaults.Default*` constants
 80  - `internal/config`: Uses `defaults.Default*` for default values
 81  - `internal/runtime`: Uses `defaults.Default*` for runtime defaults
 82  - `internal/llm`: Uses `defaults.Default*` for LLM client defaults
 83
 84### Fixed
 85
 86- `nil` pointer safety in aggregator
 87  - Added nil checks for `progressCtx` across all log statements
 88  - `FeedAggregator` now works correctly when progress context is nil
 89  - Tests use `NewFeedAggregatorWithProgress` with explicit progress context
 90
 91## [0.5.0] - 2026-04-22
 92
 93### Added
 94
 95- **Accurate token counting with tiktoken**
 96  - New `internal/tokeniser` package using OpenAI's tiktoken-go library
 97  - BPE tokenisation for accurate token counting vs rough character estimation
 98  - Encoding caching to avoid repeated initialisation overhead
 99  - Model-aware encoding selection (cl100k_base, o200k_base, etc.)
100  - `CountTokens()` and `CountMessagesTokens()` functions
101  - Fallback to rough estimation when tiktoken initialisation fails
102- **Concurrent feed fetching**: Feeds are now fetched concurrently using
103  `golang.org/x/sync/errgroup`
104  - Semaphore limits concurrent requests to 10 to avoid server overload
105  - Mutex-protected shared state for thread-safe article collection
106  - Partial failures don't stop other feeds from being processed
107- **Scrollable summary in TUI**
108  - Bubble Tea viewport component for browsing long summaries
109  - Keyboard navigation: j/k or arrows (scroll), Space/B (page), g/G
110  (start/end)
111  - Mouse wheel scrolling support
112  - Scroll progress indicator showing position and total lines
113
114### Changed
115
116- **Streamlined configuration management**: Viper now uses global instance with
117automatic precedence handling. Simplified configuration flow: Parse → GetViper
118→ BindCLIArgs → ViperToRuntime
119- **TUI colour scheme**: All hex colour codes replaced with ANSI 256-colour
120  palette
121
122### Fixed
123
124- Viewport now properly handles scroll events via `Update()`. Summary content
125  pre-wrapped to viewport width before display
126- Fixed string literal colour names to use actual variables
127
128## [0.4.0] - 2026-04-21
129
130### Changed 
131
132- **TUI is now fully functional and production-ready**: Replaced the
133work-in-progress TUI implementation with a robust Bubble Tea command pattern:
134  - Runtime execution now uses native `tea.Cmd` instead of manual goroutine
135  orchestration.
136  - Added real-time progress updates via `progress.Progress` interface with
137  stage, substage, and article count messages.
138  - TUI displays a spinner, elapsed time, and final summary inline.
139  - Removed signal handling and channel synchronisation in favour of Bubble
140  Tea's built-in lifecycle.
141  - `TUIProgress` now only sends messages to an existing program, simplifying
142  integration.
143  - Both verbose mode (`SimpleLogger`) and TUI mode share the same
144  `progress.Progress` interface.
145
146## [0.3.0] - 2026-04-21
147
148### Added
149
150- **Configuration file support via TOML** (XDG-compliant and cross-platform)
151  - Load settings from `~/.config/llm_aggregator/config.toml`
152  - Supports all aggregation, API, and output options
153  - Example configuration file at `configs/config.example.toml`
154- **Environment variable support with `LLM_AGGREGATOR_` prefix**
155  - All configuration options can be set via environment variables
156  - Precedence order: CLI arguments > environment variables > config file >
157  built‑in defaults
158- New `internal/config` package with Viper integration
159  - `Load()` and `Save()` methods for configuration management
160  - `GetConfigPath()` for XDG‑compliant config file location
161  - `ConfigExists()` to check for existing configuration
162- Comprehensive unit tests for configuration loading, saving, and environment
163  variable precedence (`internal/config/config_test.go`)
164
165### Changed
166
167- **Environment variable renamed:** `DEEPSEEK_API_KEY` →
168  `LLM_AGGREGATOR_API_KEY`
169- **CLI help text** now reflects the new API key environment variable and
170  general configuration options
171- **Command‑line argument precedence** implemented in `cmd/llm_aggregator.go`
172via `applyConfiguration()`
173- Updated all help text, documentation, and code references
174- `README.md` completely revised with a new "Configuration" section
175
176### Removed
177
178- Global `QuietMode` and `VerboseMode` variables from
179`internal/config/config.go`. Replaced by runtime‑specific configuration in
180`runtime.Runtime` and CLI arguments
181- Hardcoded system prompt in `runtime.NewRuntime()`. Now left empty, allowing
182configuration or client default to be used
183- CLI flags now correctly override config file and environment variables
184
185## [0.2.0] - 2026-04-21
186
187### Changed
188
189- Moved all progress and status messages behind `-v/--verbose` flag
190- Default CLI mode is now silent except for final output and errors
191- Components (aggregator, processor, LLM client) use logger interface
192controlled by verbose flag
193
194## [0.1.0] - 2026-04-21
195
196### Added
197
198- Complete translation from Python prototype to Go implementation
199- RSS/Atom feed parsing using `gofeed` library
200- DeepSeek API integration via `openai-go` client, configured for `/chat/completions` endpoint
201- Content filtering and processing with keyword‑based include/exclude, date filtering, and sorting
202- Command‑line interface using `go‑arg` with automatic help and version flags
203- Multiple output formats: plain text, GitHub‑flavoured markdown, and JSON
204- Terminal user interface (TUI) built with `bubbletea` featuring:
205  - Animated progress bar with gradient colours (`#FF7CCB` to `#FDFF8C`)
206  - Live article counters (aggregated/processed)
207  - Elapsed time display
208  - Coloured status indicators using `lipgloss` styling
209  - Keyboard controls (q/Ctrl+C to quit)
210- Web content extraction fallback using `goquery` when feed descriptions are minimal
211- Configurable limits: articles per feed, maximum age, total articles
212- Token estimation and API usage logging
213- Environment variable support (`LLM_AGGREGATOR_API_KEY`) for authentication
214- Example feeds file with technology, programming, and free software sources
215
216### Changed
217
218- Program structure organised into standard Go layout: `cmd/`, `internal/`, `pkg/`
219- English spelling conventions maintained throughout (colour, initialise, summarise)
220- Error handling improved with specific messages for common API failures (invalid key, rate limits, etc.)
221- Progress reporting unified through a `ProgressContext` interface for both TUI and CLI modes
222
223### Fixed
224
225- Initial API endpoint mismatch (using `/responses` instead of `/chat/completions`)
226- Nil pointer dereferences when handling optional feed metadata (author, dates)
227- String repetition syntax errors in Go (replaced `"="*80` with `strings.Repeat`)
228- CLI help/version flag handling to show information without requiring other arguments
229- Type compatibility issues with `openai-go` v3 API (message parameters, token usage fields)