Phase 2: Developer Experience (v0.2.0)#27
Merged
Merged
Conversation
EventBus with exact/prefix/catch-all subscriptions, FelixEvent frozen dataclass, EventType enum (agent.*, workflow.*, task.*, message.*, stream.*, spawn.*), and EventEmitterMixin. Wired into FelixWorkflow (6 lifecycle events), LLMAgent (task start/complete), and CentralPost (lifecycle bridge). All event bus params are optional — zero overhead when not used. 35 new tests. 698 total passing. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- EventEmitterMixin now writes _event_bus to instance dict explicitly via __dict__ to prevent class-level attribute leakage between instances - Always call set_event_bus() in FelixWorkflow/LLMAgent __init__ (even when None) so the attribute exists for direct access in runner.py - Added TODO comments on placeholder EventType members not yet emitted Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
feat(events): add event system for SDK observability
- configure_logging() one-call setup for all felix_agent_sdk.* loggers - FelixLogConfig: level, format (text/json), per-subsystem overrides - JSONFormatter: structured JSON output with event metadata fields - EventLogBridge: subscribes to EventBus, auto-logs events at configurable levels (workflow events at INFO, task/agent at DEBUG) - FelixEvent.__post_init__ normalises EventType enum members to plain strings for consistent str()/logging across Python versions 16 new tests. 714 total passing. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
EventBus.enable_history() now accepts max_size (default 10,000). Oldest events are discarded when the limit is reached during emit(). Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
feat(logging): add structured logging and EventLogBridge
- StreamEvent/StreamEventType frozen dataclasses for token-level events - StreamHandler base class with dispatch routing (token/chunk/result/error) - CallbackStreamHandler for simple callable-based consumers - EventBusStreamHandler re-emits stream events onto the event bus - StreamAccumulator bridges provider StreamChunk to SDK StreamEvent, produces CompletionResult for parity with non-streaming path - LLMAgent.process_task_streaming(task, handler) — full streaming processing with same lifecycle as process_task() 23 new tests. 737 total passing. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
feat(streaming): add streaming types, handlers, and accumulator
- ConfidenceMonitor: tracks per-agent/team confidence trends, detects stagnation, recommends SPAWN/HOLD/PRUNE - ContentAnalyzer: keyword-based topic coverage gap analysis, recommends agent type to fill gaps - TeamSizeOptimizer: heuristic team size from task complexity signals - DynamicSpawner: orchestrates monitor + analyzer, creates agents via factory, connects via spoke manager, emits spawn events - WorkflowConfig: enable_dynamic_spawning and max_dynamic_agents fields - FelixWorkflow: _check_dynamic_spawning replaced with inline spawner integration, new agents join subsequent rounds 35 new tests. 772 total passing. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Eliminates triple duplication of stopword sets across compression.py, task_memory.py, and content_analyzer.py. Single frozenset in utils/text.py, aliased locally. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
feat(spawning): add dynamic confidence-driven agent spawning
- felix init <name> [--template] — scaffolds project with felix.yaml, main.py, requirements.txt, .env.example. Templates: research, analysis, review. - felix run <config.yaml> [--provider] [--verbose] — loads YAML config, resolves provider, runs workflow, prints result. - felix version — prints SDK version. - yaml_loader: parses helix presets or custom dicts, team composition, all WorkflowConfig scalars including dynamic spawning fields. - Console script entry point in pyproject.toml. 20 new tests including init/run round-trip. 792 total passing. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
feat(cli): add felix console command with init, run, version
Extract and generalize the demo helix visualizer from examples/05_deep_research_live/ into a proper SDK module at src/felix_agent_sdk/visualization/. The new module provides: - terminal.py: ANSI color primitives, cursor control, progress bars - helix_visualizer.py: HelixVisualizer class with agent registration, live update, sidebar panel, phase boundaries, and team confidence - AgentDisplayState dataclass for per-agent rendering state - Exported via top-level __init__.py as HelixVisualizer Includes 23 unit tests covering registration, update, rendering, NO_COLOR support, context manager, and dimension customization. All 686 tests pass, ruff clean. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Addresses review feedback from @CalebisGross: - Remove unused `import os` in test file (ruff F401) - Apply `ruff format` to helix_visualizer.py and test file Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Move the colour-name → ANSI-code mapping from helix_visualizer.py into terminal.py as COLOR_MAP, eliminating duplicated constants. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Wire the HelixVisualizer into all four features that landed on dev/phase-2: - Event bus: attach_event_bus() auto-registers agents on AGENT_SPAWNED, updates progress on AGENT_POSITION_UPDATED, marks DONE/FAILED - Streaming: VisualizerStreamHandler shows live token count in sidebar - Spawning: set_monitor() displays confidence trend and spawn recommendation in the footer - CLI: felix run --visualize / -V flag for live helix during workflows 31 visualizer tests pass (23 existing + 8 new). 823 total tests pass. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
feat: add reusable terminal helix visualizer module
- New examples: event_system, streaming, dynamic_spawning, structured_logging, yaml_workflow - Fix yaml_loader _parse_team .pop() mutation (use .get() + filter) - CHANGELOG.md: full v0.2.0 section covering all Phase 2 features - README: update roadmap to reflect Phase 2 completion - Version bump: 0.1.0 -> 0.2.0 Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- 06: extract lambdas to named functions for readability - 08: show gap-to-threshold instead of premature agent_id in SPAWN_TRIGGERED handler - 09: use INFO level for JSON demo (less noise than DEBUG) - 10: clarify API key requirement in README Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
feat: examples v2, version bump to 0.2.0, changelog
Member
Author
|
@CalebisGross — v0.2.0 is out. This one adds all the developer experience tooling on top of your core algorithms: event observability, structured logging, token-level streaming, confidence-driven dynamic spawning, and a CLI ( Your helix geometry, hub-spoke routing, and agent lifecycle are still the foundation everything builds on. The dynamic spawning module in particular — ConfidenceMonitor, ContentAnalyzer, DynamicSpawner — is a direct formalization of the patterns you designed in the original Felix. It's the feature that makes this framework genuinely different from everything else out there. If you want to check it out: Release: https://github.com/AppSprout-dev/felix-agent-sdk/releases/tag/v0.2.0 |
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
Phase 2 adds developer experience features to the Felix Agent SDK. Zero breaking changes to v0.1.0 API.
What's new
EventBuswith pub/sub,FelixEvent,EventTypeenum,EventEmitterMixin. Wired into FelixWorkflow, LLMAgent, CentralPost.configure_logging(), JSON/text formatters,EventLogBridgefor automatic event-to-log bridging.StreamHandler,StreamAccumulator,LLMAgent.process_task_streaming()with token-level events.ConfidenceMonitor,ContentAnalyzer,DynamicSpawner. Confidence-driven agent creation wired into the workflow round loop.felix init(scaffold projects),felix run(YAML config),felix version. Console script entry point.Stats
PRs included
Test plan
felix versionoutputs 0.2.0felix initscaffolds valid project