- Added `ExploreService` trait and `ExploreServiceImpl` struct to handle the exploration of codebase context before agent turns.
- Implemented three parallel subagents: Code Structure, Symbol Index, and Semantic Context, each with specific directives.
- Integrated the explore phase into the agent turn process, ensuring that each turn starts with a consolidated context message.
- Enhanced `spawn_agent_turn` function to include explore service wiring and context preparation.
- Implemented a code quality scanner that checks for common clean-code violations in Rust source files, including missing documentation, usage of `.unwrap()` in production code, and commented-out code.
- Introduced a commit message validator that follows the Conventional Commits specification, ensuring proper formatting and providing suggestions for invalid messages.
- Created a unified BestPracticeEngine to encapsulate the functionalities of skills, architecture audits, code quality checks, and commit message validation.
- Added tests for both the code quality scanner and commit message validator to ensure reliability and correctness.
- Created solid.md to document the SOLID principles for clean code practices.
- Created tdd.md to outline Test Driven Development principles and practices.
- Added kana-rust-backend-best-practice.md as a reference guide for building a Rust backend using Axum and SeaORM.
- Established push-flow-convention.md to enforce pre-commit and pre-push hooks with versioning rules.
- Introduced AGENTS.md to provide guidance on best practices and available commands for Kilo.
- Configured kilo.json to include new skills and agents for enhanced functionality.
- Added lefthook.yml for managing git hooks to ensure code quality and adherence to conventions.
- Add AppStateRest as the central state struct for managing TUI state.
- Implement InputState for handling user input, autocomplete, and history.
- Create MiscState to manage overlays, notifications, and editor state.
- Introduce ScrollState for viewport scrolling functionality.
- Develop TranscriptCache for efficient message rendering in the chat pane.
- Implement SimpleAgent and SimpleWorkflowEngine for agent lifecycle management.
- Add helper functions for managing effort levels and token counting.
- Organize state-related modules for better maintainability and clarity.
- Moved `AccessTier` and `SubagentEvent` enums to `zesdex_domain::subagent`.
- Consolidated workflow-related types into `zesdex_domain::workflow`.
- Updated references across the codebase to use the new domain models.
- Refactored tool execution logic to utilize a new `ToolExecutor` trait.
- Enhanced `AgentTurnService` to handle tool calls and events more effectively.
- Adjusted API handlers and state management to align with new domain structure.
- Implemented a new tool for semantic code search that indexes Rust code symbols (functions, structs, enums, traits, modules) and allows searching by name, concept, or meaning.
- Introduced a symbol index structure with methods for rebuilding the index and searching symbols.
- Added regex patterns for extracting various code symbols from Rust source files.
- Implemented scoring logic for search results based on exact matches, prefix matches, and context relevance.
- Created a web search tool that interacts with a SearXNG instance to fetch documentation and API information based on user queries.
- Added a diff preview overlay for rendering git diff output with color-coded additions and deletions in a TUI interface.
feat(shell): enhance output capturing by using threads for stdout and stderr
feat(tui): update usage widget to display token counts and provider/model information
refactor(tui): simplify status bar rendering by removing unnecessary token calculations
- Added tracing instrumentation and improved logging messages in the Pong, Todofinish, and Todowrite tools for better debugging and monitoring.
- Enhanced documentation comments for clarity on tool functionalities and workflows.
- Implemented tracing in WorkflowRun, NoteFinding, ReadFindings, and HiveMind tools to track execution phases and findings.
- Updated TUI overlays (e.g., Bash, Clear Confirm, Editor, Effort Level, Help, Key Input, Learning, Loading, MCP, Model Selector, Plan, Quit Confirm, Rewind, Settings, Todo, Usage) with debug logging to capture rendering details.
- Improved the status bar and workflow panel rendering with additional debug information.
- Added tracing to various utility functions to facilitate better performance monitoring and error tracking.
feat(bootstrap): create temporary settings and config files to prevent data loss
refactor(edit_log): switch from Vec to VecDeque for efficient memory management
fix(gateway): ensure store directories are created before starting the API server
refactor(bgbash): implement a global singleton for BashControl
feat(auth): enhance session authentication middleware to use SessionRepository
fix(edit_log_repo): update to use VecDeque for in-memory edit log storage
fix(memory_repo): add newline escaping for frontmatter fields
fix(session_lock_repo): improve error handling for lock file operations
fix(bash_tools): prevent path traversal in job_id argument
refactor(delete): enforce empty directory deletion in file system tools
fix(edit): optimize string replacement to only replace the first occurrence
fix(git_cred): improve credential management with piped input to git commands
feat(git_operator): add safety filter to block destructive git operations
fix(shell): register background jobs in Bash control
feat(spawn): add access tier specification for pipeline stages
refactor(hive_mind): run directives concurrently for improved performance
fix(auth): update refresh token verification in the refresh handler
fix(chat): optimize LLM client usage based on model matching
fix(conversations): enhance message deletion to target specific indices
feat(api): add JWT authentication middleware for all API routes
fix(state): implement refresh token verification in JwtTokenService
fix(daemon): improve usage tracking with saturating addition
fix(tui): handle compacted messages in the TUI state management
feat(tui): implement status bar with connection and turn state indicators
feat(tui): create workflow panel for agent status and progress visualization
feat(web): introduce web frontend interface with static file serving
feat(ws): add WebSocket interface for real-time communication and session management
- Removed HTTP adapter module from CMS infrastructure.
- Updated CMS infrastructure module to exclude HTTP.
- Introduced presentation layer in CMS with DTOs and handlers for REST API.
- Added command types for CMS domain operations to encapsulate input data.
- Created typed error handling for CMS presentation layer.
- Implemented handlers for CMS REST API endpoints.
- Removed HTTP DTOs and handlers from IAM infrastructure.
- Introduced command types for IAM domain operations.
- Created presentation layer in IAM with DTOs and handlers for OAuth flow.
- Implemented typed error handling for IAM presentation layer.
- Introduced `SessionId` newtype for validated session identifiers, ensuring safety against path traversal attacks.
- Updated session repository methods to accept `SessionId` instead of raw strings, enhancing type safety.
- Removed redundant error handling in repository methods by leveraging the new `Error` type from `zesdex_utils`.
- Simplified atomic JSON write operations by eliminating unnecessary error conversions.
- Enhanced integer casting with a new `CastOr` trait for safer narrowing conversions.
- Removed deprecated error handling code and consolidated error types across the codebase.
- Updated HTTP handlers to utilize the new session ID validation, improving overall robustness.
- Introduced `RepositoryError` and `ServiceError` enums in both IAM and CMS domains for better error management.
- Updated domain traits and services to return specific error types instead of `anyhow::Result`.
- Enhanced session and OAuth repository implementations to handle errors more explicitly.
- Refactored session service methods to return `Result<T, ServiceError>` for improved error handling.
- Updated HTTP handlers to utilize the new error types.
- Modified password hashing functions to run in a blocking context using `tokio::task::spawn_blocking`.
- Added tests for new error handling mechanisms and async password functions.
Extract duplicated session ID extraction + validation logic from
SessionAuthMiddleware::call() and require_session() into two shared
helper functions: extract_session_id and validate_and_build_identity.
Removes ~60 lines of duplicated code while preserving behavior:
- Both call sites now rely on the same extraction/validation path
- User-Agent default remains empty string (existing behavior unchanged)
- Error response format (401 with header/validation messages) unchanged
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Add the cast-allow block to zesdex-entities/src/lib.rs and
zesdex-utils/src/lib.rs (which lacked it), then remove from
65 sub-files across all 8 crates. Build and all 223 tests
continue to pass.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- Simplified token type assignment in OAuth service.
- Removed unused session_lock module and re-exported Session from zesdex_entities.
- Cleaned up session entity by removing unnecessary comments and code.
- Consolidated session handling in HTTP handlers for better readability.
- Improved formatting and readability in OAuth repository tests.
- Enhanced session lock repository with clearer match statements.
- Streamlined session repository error handling.
- Refined RNG tests for better clarity.
- Adjusted module visibility and organization in lib.rs.
- Updated IPC client and connection code for better error handling and clarity.
- Improved frame handling in IPC for better readability.
- Organized module imports and added test utilities for IPC.
- Enhanced database connection error handling.
- Simplified JWT token creation error handling.
- Improved password verification error handling.
- Cleaned up state management code for better readability.
- Refactored middleware for session authentication and rate limiting.
- Simplified clipboard utility for better error handling.
- Enhanced logging initialization for better error reporting.
- Improved pagination utility with clearer method annotations.
- Cleaned up sanitization functions for filenames and paths.
- Enhanced slug generation functions for better clarity and usability.
Ganti semua pemanggilan Memory::read/write/remove/list di zesdex-backend
dengan MarkdownMemoryRepository dari zesdex-cms. Hapus re-export
model::memory yang sudah tidak dipakai.
Method mapping: read -> load, write -> save, remove -> delete, list -> list.
Import trait MemoryRepository untuk method resolution.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Semua pemanggilan AppConfig::load() diganti dengan
JsonAppConfigRepository + AppConfigRepository trait.
Re-export model::app_config dihapus dari model/mod.rs.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Duplikasi verbatim dari crates/zesdex-backend/src/service/oauth/loopback.rs
ke zesdex-iam untuk sentralisasi primitif OAuth.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
- Updated markdown rendering logic to use more concise methods for obtaining vector lengths.
- Changed review status display to use the correct flag from settings.
- Cleaned up sidebar rendering code for better formatting and readability.
- Enhanced status bar rendering with improved string formatting and consistent style application.
- Refined workflow panel rendering, ensuring consistent style usage and improved readability.
- Added architecture overview and detailed documentation for backend, data, dependencies, and frontend structures.
Ditemukan reviewer whole-branch final: looks_log_shaped murni berbasis
konten (>=3 baris berpola error/warn/fail), jadi hasil grep/search yang
match ke kode error-handling ikut lolos ambang itu -- padahal
squash_log punya cap keras 20 error + 10 warning tanpa budget byte,
diam-diam membuang match yang sah di luar cap itu. Sekarang hanya tool
bash (penghasil log sungguhan) yang boleh lewat squash_log; tool lain
yang kebetulan konten-nya mirip log jatuh ke squash_generic yang lebih
longgar (head/tail + budget byte). Tambah test regresi yang membedakan
kedua jalur lewat retensi baris terakhir.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Perbaikan gaya murni (format! di-append ke String -> write!, syntax
perbandingan yang lebih jelas, backtick di doc comment) tanpa
mengubah perilaku -- semua test tetap hijau.
Off by default, diaktifkan lewat settings.json (belum ada UI toggle —
sama seperti review_enabled/session_archive_enabled/lsp_auto_provision
yang juga cuma bisa diedit manual hari ini). Saat aktif, system prompt
diberi instruksi menulis ringkas, dengan pengecualian eksplisit untuk
konfirmasi operasi destruktif dan peringatan keamanan yang tetap harus
detail penuh.
Status bar sekarang memakai penghitungan token yang sama persis dengan
compaction (bukan heuristik /4 terpisah), dan selalu menampilkan angka
context window nyata alih-alih '?' saat model role tidak override
context_window secara eksplisit — konsisten dengan fallback yang
sudah dipakai compaction sendiri.
Salinan yang dikirim ke UI (TurnEvent::ToolResult) tetap utuh; hanya
salinan yang masuk riwayat percakapan (dikirim ke LLM) yang dikompres,
supaya user tetap melihat output tool apa adanya.
Masih menyebut shortsend padahal modul itu sudah dihapus dan diganti
context::dedup/context::shaping di commit sebelumnya.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Loop auto-compaction sekarang selalu menjalankan dedup tiap iterasi
lalu shaping lewat context::, menggantikan shortsend:: yang dihapus.
Action::Compact tadinya berjalan sinkron dan selalu client: None,
sehingga hasil compact manual tidak pernah diringkas LLM (beda dengan
compaction otomatis di tengah turn). Sekarang /compact jalan di
thread background seperti spawn_turn, sehingga bisa memanggil LLM
untuk meringkas riwayat yang dibuang — perilaku manual dan otomatis
jadi setara.
Ekstrak resolve_llm_client_config() dari spawn_turn supaya logika
resolusi api_key/model/base_url tidak dua kali.
Ditemukan implementer Task 5: fixture pesan pendek (~8 token nyata
lewat tiktoken untuk 20 pesan = ~160 token) tidak pernah melebihi
target 700 token (70% dari max_wire_tokens=1000), jadi force=true pun
tidak pernah men-drop satu pesan pun -- test placeholder-summary dan
keeps-most-recent lulus secara vakum tanpa benar-benar menguji jalur
drop. Diverifikasi ulang dengan context::tokens::count_tokens nyata:
fixture baru (20 pesan ~49 token = ~980 token total) melebihi target
700 dengan nyaman.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Perilaku should_shape/shape_messages tidak berubah, hanya sumber
penghitungan token yang sekarang lewat context::tokens (tiktoken-rs)
menggantikan heuristik char/3 bawaannya sendiri.
Kompresi per-jenis-konten (JSON: pertahankan struktur & value pendek/
entropi tinggi, buang value panjang bertele-tele; log: simpan baris
error/warning berskor tertinggi + konteks sekitarnya; generic: potong
importance-ranked) untuk hasil tool di atas 1.5KB. Tool read dikecualikan
total karena isinya harus tetap byte-exact untuk edit selanjutnya.
Ditemukan lewat trace manual (bukan implementer): fixture lama cuma
punya 1 baris berpola error, di bawah ambang looks_log_shaped (>=3),
jadi tanpa sadar selalu jatuh ke squash_generic -- test tetap lulus
tapi tidak pernah menguji logika scoring/windowing squash_log sama
sekali. Tambah baris error/warning lagi supaya jalur squash_log
benar-benar terpakai.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Ditemukan implementer Task 4 (percobaan kedua, lagi-lagi BLOCKED
sebelum commit apapun): test sebelumnya memakai string sama yang
panjang+entropi-rendah di semua 4 elemen array, jadi elemen index
0-2 pun ikut ter-elide oleh aturan panjang/entropi normal -- tidak
benar-benar menguji efek posisi array. Ganti fixture pakai string
berbentuk UUID (tanpa spasi, entropi tinggi) yang lolos aturan normal
di posisi manapun, supaya index 3 yang di-force-elide walau
identifier-shaped benar-benar membuktikan aturan "past index 3
regardless of length/entropy".
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Ditemukan implementer Task 4 sebelum commit apapun (BLOCKED, bukan
kode salah): entropi Shannon mentah per-karakter tidak membedakan
prosa dari identifier acak — prosa berulang skor ~3.89 bit/char,
lebih tinggi dari UUID (~3.39). Tambah syarat "tanpa spasi" sebelum
cek entropi (meniru pre-filter headroom sendiri), turunkan ambang ke
3.0 pada skala mentah. Fixture test array JSON juga diperbesar
(repeat 5 -> 8) karena sebelumnya tidak pernah melewati SQUASH_FLOOR_BYTES
yang diasumsikan test itu sendiri.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Panggilan tool read-only (read, grep, glob, dst) dengan argumen persis
sama menyisakan satu salinan penuh saja di context; entri lama diganti
placeholder tapi tool-call-nya sendiri tetap terlihat di riwayat. Tool
bersifat mutasi (write/edit/bash/dll) tidak pernah disentuh.
Jadikan tool_scope::READ_TOOLS pub supaya jadi satu-satunya sumber
klasifikasi read-only, dipakai ulang bukan didaftar dua kali.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Satukan tiga salinan logika resolusi context_window (Action::Compact,
spawn_turn, status bar) yang sempat melenceng satu sama lain.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Ganti tiga heuristik char-count (/3 di shortsend, /4 di loop turn, /4
di status bar) yang saling tidak konsisten dengan satu BPE tokenizer
nyata. tiktoken-rs membundel vocab lewat include_str! saat build, jadi
tidak ada akses jaringan saat runtime.
Fixture pendek sebelumnya (~8 token nyata per pesan lewat tiktoken)
tidak pernah melebihi target 700 token, jadi shape_messages tidak
pernah men-drop satu pesan pun -- satu test gagal, satu test lain
lulus secara vakum. Pakai fixture lebih panjang (~49 token/pesan)
yang diverifikasi melebihi target dengan nyaman.
9 task: tokens.rs (tiktoken-rs), window.rs, dedup.rs, squash.rs,
shaping.rs (port shortsend), gabungan cutover+fix /compact manual,
wiring squash, status bar, dan mode ringkas opsional. Setiap task
diverifikasi dengan cargo build/test nyata, bukan -D warnings (yang
sudah merah di main karena warning pre-existing di file lain).
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Facade cuma dipakai generically oleh satu caller (Action::Compact);
auto-loop tetap butuh kontrol per-stage sendiri. Selaras dengan prinsip
"No DI" di CLAUDE.md. dedup::collapse juga diubah mengembalikan
(Vec<ChatMessage>, bool) supaya caller tahu ada perubahan tanpa perlu
ChatMessage: PartialEq.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Ditemukan saat menulis plan: kompresi JSON pada hasil read akan merusak
byte-exactness yang dibutuhkan untuk edit selanjutnya jika file yang
dibaca kebetulan berformat JSON (package.json, tsconfig.json, dst).
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Rencana pengganti shortsend.rs dengan modul context/ (dedup, squash,
shaping, tokens, window) plus mode ringkas opsional, disusun dari studi
teknik rtk-ai/rtk, headroomlabs-ai/headroom, dan JuliusBrussee/caveman
(ide saja, tanpa menyalin kode).
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
AppStateRest::new() men-spawn thread ignore::Walk untuk mention_index
tanpa syarat, padahal mode --attach cuma dipakai untuk render lokal —
handle_key dan logika mention berjalan di sisi daemon lewat IPC, jadi
index di client attach tidak pernah dipakai. Ini bikin setiap
--attach melakukan full workspace walk (sampai 50.000 entry) sia-sia.
Pindahkan thread-spawn itu ke method terpisah
spawn_mention_index_build(), dipanggil eksplisit dari
run_single_process() dan run_daemon() setelah AppStateRest::new(),
tapi sengaja tidak dipanggil dari run_attach().
Cursor bisa berpindah (Left/Right) tanpa menutup dropdown mention,
sehingga mention_start jadi stale relatif ke cursor saat Enter
ditekan. select_autocomplete() lalu memanggil replace_range dengan
start > end dan panic (crash seluruh TUI, termasuk daemon). Tambah
guard: jika cursor < mention_start atau mention_start > buffer.len(),
tutup dropdown dan kembalikan false alih-alih menyambung range yang
tidak valid. Tambah regression test yang mereproduksi skenario ini.
asepharyana