add self-correcting JSON in case of invalid structure; update README (#…

…65)

feat: set DOCKER_MODE=1 in Dockerfile template

Add DOCKER_MODE=1 environment variable to the Dockerfile template to
ensure that containers built from this template will skip the
~/.kdeps.pkl config file lookup. This aligns with the Docker mode
behavior and prevents unnecessary file system lookups in containerized
environments.

The environment variable is set alongside other Docker-specific
environment variables (SCHEMA_VERSION, OLLAMA_HOST, KDEPS_HOST) for
consistency.

upgrade schema to v0.4.1-dev

Fix: Allow winget installation to fail without blocking release

The winget installation step was failing due to VCLibs version conflict
(newer version already installed), which blocked the entire release pipeline.
Since the winget test is disabled anyway, we can safely continue on error.

add more coverage tests; trigger multiplatform test on PR create (#62)

* add more coverage tests; trigger multiplatform test on PR create

* additional code-coverage tests

Offline part 2 (#61)

* update schema to 0.2.44

* migrate to schema 0.2.44 using pkl-go 0.11.1 and pkl 0.29.0

* cleanup and remove workaround code

* use pkl assets imports on dockermode and test; upgrade to schema 0.2.50

* use pkl 0.29.0 on Dockerfile

* add default Nat HostIP configurations

* pkl-go 0.11.1 fixes for kdeps cfg

* pkl-go 0.11.1 fixes for kdeps action types

* add ipv6 loopback to default NAT config

* add progress bar when downloading cache

* revert resources.go changes for pkl-go 0.11.1

* upgrade to schema v0.3.0

* update packages

* upgrade to schema v0.3.1-dev

* add version suffix support

* migrate to pkl-go 0.11.1

* fix ollama binary check

* fix 404 error test

* schema 0.3.1-dev and pkl-go 0.11.1 migration

* golangci fixes 1

* golangci fixes 2

* golangci fixes 3

* golangci fixes 4

* download ollama model when not found

* feat: Enhanced Ollama model management and stability improvements

- Add proactive model availability checking before LLM operations
- Improve error detection to focus on specific 'try pulling it first' errors
- Implement automatic model syncing from /root/.ollama/models/ to /models/ after successful pulls
- Enhance rsync commands with --progress flag for better visibility
- Fix nil pointer dereferences in timestamps.go for improved stability
- Update schema version handling to use dynamic versions instead of hardcoded values
- Set GIN web framework to release mode to eliminate debug warnings
- Expand test coverage with comprehensive error detection tests
- Add detailed logging for model operations and sync processes

This commit significantly improves the robustness of Ollama model handling,
ensures model persistence across container restarts, and provides better
user feedback during model operations.

* sync ollama models to persistent storage, and websocket fixes

* update go modules

* update schema to 0.3.1-dev

* upgrade GHA to pkl version 0.29.0

* fix issue when no models are defined to pull

* fix test to properly handle binary availabbilty errors

All reader DB uses :memory: except for MemoryReader

automatically creates shared .kdeps folder on docker environment if m…

…ounted shared volumes are not available

HOTFIX: (Offline-first) don't block ollama pull when OfflineMode is e…

…nabled but there's no internet.

added full LLM offline-mode (#58)