docs: API keys, providers, and fallback configuration#20
Merged
kargarisaac merged 12 commits intomainfrom Mar 29, 2026
Merged
Conversation
- Explain .env file loading (python-dotenv, from project root) - Add provider table with env vars and config values - Document fallback_models configuration with example - Remove stale ResponsesProxy reference Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…loading - lerim init now includes LLM provider selection with API key entry - Keys saved to ~/.lerim/.env with 600 permissions - load_dotenv always checks ~/.lerim/.env (not just CWD) - Shows all 7 providers with descriptions - Preserves existing keys, supports multiple providers - Beautiful terminal UX with section headers - README updated with provider table and ~/.lerim/.env guidance Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The evals framework (benchmark runners, golden dataset comparisons, LLM-as-judge scoring, judge prompts) is optimization infrastructure, not a user-facing feature. Moving it to the private repo protects quality criteria, reward signal design, and optimization strategies. Removed: - evals/ directory (7 runners, scores, judge, configs, dataset, prompts) - tests/unit/test_eval_scores.py, tests/integration/test_eval_runners.py, tests/integration/test_judge.py - site/evals/ Renamed: - build_eval_config() → build_isolated_config() in settings.py Kept: - CHANGELOG.md historical entries (unchanged) - Integration tests that don't depend on evals (test_extract, test_summarize, test_dspy_adapters) - build_isolated_config() function body (used by lerim-cloud via editable install) 600 unit tests pass. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Updated the job claiming logic to process multiple sessions in a loop, allowing for better handling of pending jobs. - Enhanced the extraction and failure tracking by accumulating results from each batch of claimed jobs. - Improved overall efficiency by reducing the number of claims made when processing jobs. This change aims to streamline the synchronization process and improve performance in job handling.
When the LLM correctly determines a session has no extractable memories, it returns empty <primitives></primitives> tags. DSPy's XMLAdapter fails to parse the empty string as list[MemoryCandidate]. Fix with two prongs: better signature prompt guiding the model to return [], and a code-level safety net that catches the specific AdapterParseError and returns an empty list gracefully. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Providers like minimax require PascalCase (MiniMax-M2.5) while opencode_go uses lowercase. Add normalize_model_name() to provider_caps and apply it at config load and fallback spec parse time so users don't hit silent 404s from casing mismatches. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
LLMs sometimes set primitive to "pitfall", "insight", etc. instead of "learning" with kind set. Add a model_validator that auto-corrects this so downstream code doesn't reject valid candidates. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add end-to-end tracing: DSPy pipelines, OAI Agents SDK, and manual spans for sync/maintain cycles. Includes session-field scrubbing allowlist, OTel context propagation to async event loop, and OperationResult for structured span attributes and service_run details. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add /api/jobs/retry-all and /api/jobs/skip-all to handle all dead_letter queue items at once, avoiding the need to retry/skip one at a time. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Updated the max_window_tokens configuration in default.toml from 150000 to 100000 to optimize performance. Adjusted the summarization pipeline logic to account for a new prompt headroom, ensuring trace tokens fit within the adjusted limits. This change aims to enhance the efficiency of the summarization process.
Made-with: Cursor
- Add vulture to lint extras and [tool.vulture] (src/, min_confidence 100) - Register pre-push pre-commit hook - Prefix unused handler params in auth and serve signal shutdown Made-with: Cursor
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
.envfile guidance and provider table🤖 Generated with Claude Code