hablas-ai 2.2.1 → 2.2.3

package/CHANGELOG.md

@@ -1,26 +1,33 @@
 # Changelog
 
-## [2.2.1] — 2026-06-12 · Runtime Persistence and Web-Search Crash Fix
-
-### Fixed
-- Restored the missing project-state import in the interactive runtime that caused `ensureProjectStateDir is not defined` at startup.
-- Fixed the web-search follow-up crash path by restoring the missing large-web token constant and the associated output-preservation path.
-- Project session state is now saved after slash commands as well as normal turns.
-- Read-cache preservation is now actually written back after successful `read_file` operations and invalidated when related files change.
-
-## [2.2.0] — 2026-06-12 · Stateful Runtime and Thinking Hooks
+## [2.2.3] — 2026-06-12 · Trusted Image Pipeline and UX Phase Labels
 
 ### Added
-- project-local `.hablas/session.json` persistence for workspace session state
-- larger default history and context budgets for the rebuilt runtime
-- engine hooks for model-start / model-stop so the UI can show real thinking states
+- `search_image_candidates` tool for ranking image candidates before download
+- `inspect_image` tool for checking remote or local image candidates before adoption
+- richer phase labels during tool-driven execution such as analysis, candidate selection, download, and verification
 
 ### Changed
-- direct and non-interactive sessions now use the configured `contextBudget` instead of falling back to the old small default
-- project sessions are automatically saved after turns and restored on the same workspace
-- web-search and large file context handling were tightened so useful output survives into model context better
+- strengthened the system prompt around image handling so Hablas prefers ranking and inspection before downloading assets
+- upgraded tool box styling toward a cleaner premium terminal presentation
+- download flow now attempts to resolve direct image candidates from HTML pages automatically
 
 ### Fixed
-- `LARGE_WEB_RESULT_TOKENS is not defined` crash
-- thinking hooks are now wired at the engine level instead of existing only in the UI layer
-- repeated file reads are less likely to happen because full `read_file` context is preserved and cached more consistently
+- image selection pipeline is less brute-force and less dependent on blind download retries
+- image inspection can now work on local assets as well as remote sources
+
+## [2.2.2] — 2026-06-12 · System Prompt Strength Rebuild
+
+### Changed
+- Rebuilt the single-agent system prompt at a deeper architectural level.
+- Replaced weaker phrase-driven prompt behavior with a stronger role / mission / behavior-class / tool-contract structure.
+- Strengthened the prompt across all major execution domains:
+  - product reasoning
+  - architecture reasoning
+  - implementation
+  - debugging
+  - review
+  - devops thinking
+  - web and asset acquisition
+- Strengthened tool discipline and anti-fake-completion rules.
+- Reduced dependence on brittle literal examples and pushed the prompt toward generalized intent interpretation.

package/README.md

@@ -92,12 +92,16 @@ hablas run "read package.json and tell me the version"
 ### Web and asset workflow
 Hablas now keeps the professional web tool chain active:
 - `web_search`
+- `search_image_candidates`
+- `inspect_image`
 - `scrape_url`
 - `extract_links`
 - `download_asset`
 
 This means Hablas can:
 - search the web for references and assets
+- rank image candidates before downloading
+- inspect a remote or local image candidate before adoption
 - inspect pages and extract links
 - download real images/files into the project when appropriate
 - preserve full file reads in-context so long files are not silently cut into tiny fragments that force rereads