feat: add DAG-based flow model with topological execution

[dgenio]

Summary

Implements Issue #10: Design and implement a DAG-based flow model with topological execution.

DAGFlow and DAGFlowStep are added as first-class data models alongside the existing linear Flow. The executor gains a unified dispatch path — isinstance(flow, DAGFlow) routes to a new level-by-level runner while the linear path is completely unchanged.

---

What changed

File	Change
chainweaver/flow.py	DAGFlowStep (extends FlowStep with step_id, depends_on, step_type, capability_id), DAGFlow, validate_dag_topology()
chainweaver/exceptions.py	DAGDefinitionError(flow_name, reason, detail) — reason ∈ {cycle, duplicate_step_id, unknown_dependency}
chainweaver/registry.py	AnyFlow = Flow | DAGFlow; topology validated at registration time via validate_dag_topology
chainweaver/executor.py	_execute_dag_flow() runs steps level-by-level (topological sort via graphlib.TopologicalSorter); sibling key-conflict detection guards output determinism; _execute_step reused via FlowStep proxy
chainweaver/__init__.py	Export DAGFlow, DAGFlowStep, DAGDefinitionError
tests/test_flow_execution.py	9 new DAG test classes: simple chain, single node, diamond, mixed depth, sibling conflict, level schemas, linear backward compat, missing tool, step_type
tests/test_registry.py	TestDAGFlowRegistration — 9 tests covering valid registration, coexistence with linear flows, duplicate ID, unknown dep, cycle, self-loop, overwrite, get, match-by-intent
README.md	Added DAGDefinitionError to error table; updated v0.2 roadmap checkboxes
AGENTS.md	Updated repo map; added DAGFlow/DAGFlowStep row to common-tasks table
docs/agent-context/architecture.md	Updated module-boundaries table and decision-context table (DAG execution, step_type/capability_id rationale, registration-time validation)

Why

DAG execution is ChainWeaver's core identity per the weaver-spec ("Deterministic DAG/flow orchestration"). The existing Flow model was linear-only with three # TODO (Phase 2) placeholders in flow.py.

The step_type + capability_id fields are zero-cost forward-compat slots for weaver-spec Invariant I-07 (kernel delegation). Only "tool" is executed today; "capability" is reserved for the future KernelBackedExecutor.

---

Testing performed

ruff check chainweaver/ tests/ examples/      → All checks passed
ruff format --check chainweaver/ tests/ examples/ → 17 files already formatted
python -m mypy chainweaver/                   → Success: no issues found in 8 source files
python -m pytest tests/ -v                   → 112 passed in 0.33s

---

Executor invariants preserved

• No LLM calls in executor.py
• No network I/O in executor.py
• No randomness in executor.py

---

Docs and AI-agent instruction updates

• README.md — error table + roadmap updated
• AGENTS.md — repo map + common-tasks table updated
• docs/agent-context/architecture.md — module boundaries and decision context updated
• .github/copilot-instructions.md — no change needed (routes to AGENTS.md/docs)
• .github/instructions/chainweaver.instructions.md — no change needed
• .github/instructions/tests.instructions.md — no change needed

Checklist

• All four validation commands pass locally
• Both success and error paths are tested
• __init__.py __all__ updated for all new public symbols
• Linear Flow execution path completely unaffected
• No new contradictions introduced between docs
• AGENTS.md updated (architecture changed)
• One logical change in this PR

[Copilot]

Pull request overview

Adds first-class DAG support to ChainWeaver flows by introducing new DAG data models, validating DAG topology, and executing DAGs level-by-level while preserving existing linear Flow behavior.

Changes:

• Introduces DAGFlow / DAGFlowStep models plus validate_dag_topology() for cycle/ID/dep validation.
• Extends FlowRegistry and FlowExecutor to accept and execute DAG flows (with sibling output key conflict detection).
• Adds extensive DAG-focused test coverage and updates public docs/API exports.

Reviewed changes

Copilot reviewed 10 out of 10 changed files in this pull request and generated 6 comments.

Show a summary per file

File	Description
chainweaver/flow.py	Adds DAG models and topology validation helper.
chainweaver/executor.py	Adds DAG dispatch + level-by-level DAG execution path.
chainweaver/registry.py	Accepts Flow | DAGFlow and validates DAG topology at registration.
chainweaver/exceptions.py	Adds DAGDefinitionError.
chainweaver/__init__.py	Exports new DAG symbols and error type.
tests/test_flow_execution.py	Adds DAG execution tests and keeps linear regression coverage.
tests/test_registry.py	Adds DAG registration/validation tests.
README.md	Documents new error and updates roadmap items.
AGENTS.md	Updates repo map and task guidance for new DAG fields.
docs/agent-context/architecture.md	Updates module boundaries/decision context for DAG design.