DocForge2/docs/COMPATIBILITY.md

DocForge compatibility contract

Milestone 0 establishes DocForge2 as the successor repository without renaming or replacing the working DocForge interfaces. Compatibility changes require an explicit decision, a contract-test update, and migration guidance. DocForge 2.0.0 preserves that baseline and the complete 1.4 adapter, rendering, recovery, and release surfaces recorded below.

The compatibility gate is:

make contract

The complete repository gate is:

make gate

Milestone 5 also maintains make compatibility-m5 for the frozen public, adapter, policy, projection, rendering, and retrieval matrix. make release-gate aggregates that matrix with migration, concurrency, recovery, task-evidence, adoption, version, artifact, secret-scan, browser, and benchmark proofs.

Distribution and Python imports

The Python distribution and import package remain docforge.

The installed executable names remain:

• docforge
• docforge-mcp
• docforge-viewer-manager

The top-level imports recorded by docforge.__all__ remain supported. The documented adapter, model, index, rendering, application, and MCP factory names imported from these submodules also remain supported:

• docforge.adapter_contract
• docforge.adapter_sdk
• docforge.adapter_launcher
• docforge.application
• docforge.client_config
• docforge.index
• docforge.mcp_server
• docforge.models
• docforge.policy
• docforge.render_contract
• docforge.reference_config
• docforge.reference_mcp

Names beginning with an underscore are implementation details. New public names may be added without breaking this contract.

The repository Python, JavaScript/TypeScript, and C++ adapters are supported reference implementations. Their documented configuration, evidence limits, and unsupported-fact reports are compatibility surfaces; their internal parser helpers are not adapter-authoring imports.

Milestone 4 adds three version-1 schemas without changing existing descriptor or result schemas:

• schemas/adapter-launcher.schema.json
• schemas/adapter-client-configuration.schema.json
• schemas/reference-adapter.schema.json

CLI and MCP surfaces

Existing docforge command names and arguments remain supported. Existing docforge-mcp tool names and arguments remain supported. Additive commands, tools, and response fields are allowed. Removing or changing an existing name, required argument, stable error code, or safety boundary requires an explicit compatibility decision.

Version 2.0.0 comes from one docforge._version authority. The four maintained executable surfaces report docforge 2.0.0, docforge-mcp 2.0.0, python -m docforge.reference_mcp 2.0.0, and docforge-viewer-manager 2.0.0 for --version. Generated generic and adapter client configurations include and hash-bind the same docforge_version.

MCP results retain:

• A structured status.
• Project and source identity when available.
• Stable structured domain errors.
• A bounded content warning.
• Staleness information.
• The configured output-size limit.

The result schema describes the common envelope. Operation-specific fields are additive and remain bounded by the configured tool-output limit.

The following Milestone 2 CLI additions do not change existing command signatures:

• docforge configure codex|claude|openclaw --project ROOT
• docforge doctor --client codex|claude|openclaw

Configuration output is a new version-1 machine-local contract. It preserves the docforge package and executable names and emits the existing docforge.mcp_server module entrypoint. Existing hand-written client configurations remain valid and are never rewritten automatically. Doctor is inspection-only and does not become a hidden bootstrap, synchronization, or migration path.

The following Milestone 3 CLI additions are also additive:

• docforge graph-plan VIEW_ID
• docforge graph-render VIEW_ID
• docforge graph-render-status [VIEW_ID]
• --manual-render-policy auto|explicit|disabled
• --portable-graph-policy explicit|disabled
• --live-viewer-policy on-demand|disabled

MCP adds the read-only docforge_graph_plan and docforge_graph_render_status tools. Portable graph publication remains an explicit local CLI integration action. Existing manual render, preview, visualization, and status names remain supported.

Versioned data contracts

Milestone 0 preserves:

• Project descriptor schema version 1.
• Node schema version 1.
• Edge schema version 1.
• Changeset schema version 1.
• Result-envelope schema version 1.
• SQLite index schema version 3. Version 2 indexes remain disposable and automatically rebuild; version 3 adds a source-ordered incoming-edge index for bounded impact traversal.
• Index-attestation schema version 1.
• Incremental extraction-cache schema version 1.
• Effective process-policy schema version 1. The project descriptor remains schema version 1; machine-specific capability selection is a startup binding, not canonical project content.
• Read-pagination schema version 1. Existing tool names and required arguments are unchanged. Context and changeset MCP reads accept optional limits and opaque generation-bound cursors. Direct Python changeset methods and the ordinary CLI context command retain full legacy results when pagination is not requested.
• Latest-generation-diff receipt schema version 1. The additive generation-diff CLI command and docforge_get_generation_diff MCP read accept only optional pagination fields. They record one primary-graph transition and do not create a history store or expose Logic details.
• Latest-generation-diff page schema version 1. Pages use one top-level pagination object and a nested receipt_header. stored_receipt_hash names the complete stored receipt. Opaque cursors may be restarted after a server or receipt change and are not durable public identifiers.
• Manual render-plan schema version 1.
• Graph view-plan schema version 1.
• Projection-package schema version 1.
• Projection-receipt schema version 1.
• Independent projection-policy schema version 2. Effective policy version 1 remains frozen.

Indexes, attestations, extraction caches, previews, and rendered artifacts are disposable. A schema change may rebuild them. Canonical project content and stored proposals may not be silently rewritten to satisfy a new implementation.

Adapter compatibility

An adapter implementing only:

load_projection()

remains first-class. Incremental manifests, source extraction, deterministic assembly, Logic projection, and proposal or application support are optional capabilities. Incremental adapters must retain load_projection() as their independent clean-build and equivalence oracle.

Project adapters remain explicitly composed. Generic DocForge does not discover arbitrary adapter modules or choose a project globally.

The supported generation-diff Python boundary is ProjectIndex.generation_diff(). Helpers in the docforge.generation_diff module implement the disposable publication contract and are internal; they are not frozen as adapter-authoring imports.

Preserved no-AST binding

docforge-mcp --project-root /project --no-ast is a stable shorthand for the preserve-no-ast binding policy.

The binding:

• Keeps one-method complete-projection adapters working.
• Keeps non-AST incremental fingerprinting and caching working.
• Rejects nonempty function-Logic publication.
• Rejects a pre-existing index containing function Logic.
• Blocks docforge_get_logic.
• Prevents the live viewer from pinning an index containing Logic.
• Applies the same restriction during hash-bound canonical-application refresh.
• Reports the effective policy through bootstrap and contract results.

The legacy adapter_policy payload and error codes remain unchanged. The version-1 effective_policy is additive and makes precedence, capability mode, render behavior, blocked tools, and prohibitions machine-readable.

DocForge does not inspect arbitrary adapter source to prove which parser implementation it uses. The no-AST binding is an owner-selected process policy backed by Logic publication and retrieval enforcement. It is not a filesystem sandbox and cannot stop an unrelated process with repository write access from changing adapter code.

Changesets and application

The following guarantees remain stable:

1. Registration writes one complete proposal atomically.
2. Proposal identity includes its project, root, base revision, canonical source hash, writer, and ordered operations.
3. Validation and diff inspection precede application.
4. Append, rebase, abandonment, and application use exact current hashes.
5. Stale, conflicting, unauthorized, unsafe, or invalid proposals fail closed.
6. Canonical application is absent unless one startup-bound applier is configured.
7. Derived refresh failures produce an explicit degraded receipt after canonical application. They do not make an applied proposal safe to apply twice.
8. Generic canonical publication compares exact target identity at the commit boundary. Concurrent create, update, and delete mutations fail closed, roll back when exact state remains provable, or retain recovery evidence without overwriting foreign data.
9. Per-file publication is atomic and in-process rollback covers earlier publications, but canonical application has no process-death journal and does not promise multi-file crash atomicity.
10. Cleanup degradation after semantic commit closes the proposal as applied and persists bounded application_recovery lifecycle metadata instead of returning a retryable ordinary failure.

Rendering and visualization

The generic_html renderer remains the supported version-1 manual projection. Its public GenericHtmlRenderer.prepare() signature, renderer identity, frozen alpha bytes, confined paths, raw-HTML suppression, fixed template tokens, deterministic identities, atomic replacement, and side-effect-free status remain compatible. It now delegates through a versioned manual plan, immutable package, and independent renderer.

The portable_graph_html renderer and graph_render descriptor table are additive. Manual and portable graph declarations, plans, policies, publication receipts, and status remain separate. The portable renderer does not replace the existing live viewer or docforge_visualize.

Existing project descriptors may retain any positive max_render_bytes accepted by schema version

1. A value above 20,000,000 bytes does not make the descriptor invalid, and a smaller actual artifact still renders. Actual detached worker transfer is a separate fixed 20,000,000-byte runtime boundary.

The live graph viewer remains a read-only consumer of a generation-pinned validated index. It does not become project authority or MCP retrieval authority. Source reads use the pinned index generation instead of reopening mutable canonical files behind that generation.

Task-context compatibility

docforge_get_task_context is an additive MCP read tool. The legacy docforge_get_context signature, profile compiler, direct Python results, and custom three-argument context-provider contract remain unchanged.

The new ContextCapsuleV1 and RetrievalPlanV1 types live in the public docforge.retrieval submodule. Version 1 guarantees:

• A closed task-kind vocabulary and core-derived plan. Callers cannot inject arbitrary operations, SQL, paths, relations, or Logic requests.
• One immutable index transaction and one exact project, adapter, revision, source, policy, request, plan, collection, and capsule identity.
• Deterministic bounded focus, traversal, hydration, token accounting, response packing, and continuation, with fixed version-1 ceilings of 1,000 evidence items, 100,000 candidate edges, and 10,000 task-query characters.
• Raw preservation of project-owned relation names. Only the documented versioned alias map gains task semantics; all other relations remain unclassified.
• Separate missing, incomplete, blocked, and provenance-limitation evidence.
• No-AST bindings retain task context but never add a Logic retrieval step or weaken the existing Logic prohibition.

An integration that replaces the legacy context provider does not silently receive the core task planner. Version 1 has no custom task-planner protocol. The task-context tool remains registered for additive name compatibility but returns task_context_unavailable without loading or synchronizing the custom projection.

The exact version-1 relation aliases are frozen by the MCP contract and repository contract tests. Changing an alias category requires a new planner version; it is not a silent implementation detail.

Milestone 3 adds ManualRenderPlanV1, GraphViewPlanV1, projection package and receipt version 1, and projection policy version 2. These are additive submodule and schema contracts. They do not change the legacy task-context, adapter, changeset, or effective-policy contracts described above.

Milestone 4 adds docforge.adapter_sdk, the fixed reference configuration and read-only MCP binding, immutable adapter launchers, and generated adapter client fragments. A legacy adapter with only load_projection() remains first-class and need not adopt incremental extraction, Logic, a reference configuration, or launcher metadata.

Safety boundary

DocForge remains bound to one explicit project root. It rejects absolute paths, root escapes, and symbolic-link escapes. Documentation text remains untrusted data. Normal MCP operation exposes no arbitrary filesystem access, renderer execution, shell command, Git mutation, deployment, publication, or project switching.

DocForge is not a filesystem sandbox. Mode-0700 canonical transaction directories protect against other users and ordinary path access; deliberate arbitrary tampering by another process with the same operating-system UID is outside the compatibility boundary.

The historical v1.0.0 release carried distribution metadata 1.0.0 while its module and MCP runtime reported 0.15.0. Version 2.0.0 records that inherited mismatch in its maintained migration proof and resolves current identity through one authority. See migrating from v1.

Recorded weaknesses, not compatibility promises

Milestone 0 records rather than redesigns these areas:

• Generic warm reads still repeat whole-project discovery, parsing, and validation.
• The base wheel intentionally omits Tree-sitter. JavaScript, TypeScript, and C++ syntax evidence requires the matching docforge[javascript], docforge[typescript], or docforge[cpp] extra. Python reference evidence uses the standard library and remains available in the base wheel.
• One individually oversized context entry is represented as explicit bounded omission evidence; callers use targeted retrieval for that node.
• One individually oversized changeset diff is transported as reconstructable canonical-JSON chunks. Cursors are corruption-detecting read tokens, not authenticated authorization tokens.
• Production fragment validation is currently slower than forced-full rendering at the maintained 1,000-page fixture. Full rendering remains the equivalence and recovery oracle.
• Remote render services, render farms, third-party renderer ecosystems, and a separate render MCP remain deferred.
• DocForge2 does not self-host its bootstrap documentation.