Mellea-fy Dialect Documentation¶

Per-runtime mapping rules consumed by Steps 0, 1, and 2 of the mellea-fy workflow:

Step 0 (classification): the source-runtime axis 4 selects which dialect applies.
Steps 1a/1b (inventory): the dialect doc lists the files to read and the roles they play.
Step 2 (mapping): the dialect doc supplies overrides that take precedence over the general primitive-mapping table.

Naming convention¶

Each file is <runtime>.md where <runtime> matches the value of classification.json:source_runtime (and the entries in the Step 0 axis-4 table in mellea-fy-classify.md).

Supported runtimes (this directory)¶

Runtime	File	Status
`agent_skills_std`	`agent-skills-std.md`	Mature
`claude_code`	`claude-code.md`	Mature
`openclaw`	`openclaw.md`	Mature
`crewai`	`crewai.md`	Mature
`langgraph`	`langgraph.md`	Mature
`letta`	`letta.md`	Mature
`hybrid`	`hybrid.md`	Mature (rules for ambiguous classifications)

Stubbed runtimes (not yet shipped here)¶

The following are recognised by the Step 0 classifier but do not yet have shipping dialect docs in this directory; the compiler falls back to the generic inventory and mapping rules for them, with a warning surfaced in the run's mapping_report.md:

autogen
openai_agents_sdk
smolagents

What if my runtime is not listed at all?

The Step 0 classifier returns unknown. The compiler proceeds with the generic inventory pass and the general mapping table; the package will be generated but without runtime-specific knowledge. Open a discussion if your runtime is widely used and you want a dedicated dialect doc.