feat: outcomes-first protection tools + MCP annotations, resources, i…#28
Open
grantgumina wants to merge 1 commit into
Open
feat: outcomes-first protection tools + MCP annotations, resources, i…#28grantgumina wants to merge 1 commit into
grantgumina wants to merge 1 commit into
Conversation
…nstructions Make the server speak in outcomes rather than 1:1 REST objects so an LLM (incl. autonomous backup agents) states a goal once instead of being walked through a multi-step "wizard". Layers 1-2 of the agreed plan; all read-only, additive, no new write paths. Layer 1 (conversational ergonomics): - Per-tool MCP ToolAnnotations via a registration table (src/tool_annotations.py): reads -> readOnlyHint; kill_job/suspend_job/disable_schedule -> destructiveHint; resume/resubmit -> idempotentHint; human-friendly titles. Clients can now self-serve lookups and gate ONLY irreversible actions. - Rich SERVER_INSTRUCTIONS operator persona (src/config.py) replacing the one-sentence prompt: resolve IDs yourself, report outcomes, be honest about limits, confirm destructive actions, stop on zero-match. - Read-first resources (src/resources.py): commvault://glossary, capabilities (honest about what it cannot do), regions (live from storage pools), objectives. Layer 2 (goal-level tools, src/tools/protection_tools.py, all preview/read-only): - explain_objective: expand 3-2-1/2-copy/1-copy into concrete copies; badge is DERIVED from structure, not asserted. - resolve_scope: collapse the client->id->subclient chain into one call; supports client-group/name/type and REFUSES tag/resource-region selection explicitly rather than guessing; empty match is reported, never silently treated as all. - plan_protection: compile scope + objective into one reviewed plan-of-record with region-matched pools, open questions, and a deterministic plan_token. PREVIEW ONLY -- creates nothing (no provisioning endpoints exist yet). Also folds in the earlier "look up for jobs jobs" -> "jobs" docstring typo fix. Verified offline against fastmcp 3.3.1: 58 tools register with correct annotations, 4 resources resolve, and plan_protection returns a complete proposal end-to-end while mutating nothing. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
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.
Made some edits to the MCP implementation here. Goal was to make the server speak in outcomes rather than 1:1 REST objects so an LLM (incl. autonomous backup agents) states a goal once instead of being walked through a multi-step "wizard".
Changes
Also folds in the earlier "look up for jobs jobs" -> "jobs" docstring typo fix.
Verified offline against fastmcp 3.3.1: 58 tools register with correct annotations, 4 resources resolve, and plan_protection returns a complete proposal end-to-end while mutating nothing.