Merge pull request #205 from CodeForPhilly/develop

Release: v0.6.0

Author: themightychris

.env.example

@@ -17,12 +17,11 @@ NODE_ENV=development
 # Public gitsheets data repo
 # ---------------------------------------------------------------------------
 
-# Path to the codeforphilly-data working tree. Relative paths are resolved
-# from the API workspace dir (`apps/api/`) — prefer absolute paths in dev.
-# Bootstrap a fresh dev data repo with the sheet configs the API needs:
-#   npm run -w apps/api script:setup-dev-data
-# (Uses this CFP_DATA_REPO_PATH; creates the directory + git repo + .gitsheets/<sheet>.toml.)
-# In production, point at a clone of https://github.com/CodeForPhilly/codeforphilly-data.
+# Path to a BARE clone of codeforphilly-data (the API refuses a non-bare
+# clone). Relative paths are resolved from the API workspace dir
+# (`apps/api/`) — prefer absolute paths in dev. For a small seed:
+#   git clone --bare --branch fixture https://github.com/CodeForPhilly/codeforphilly-data.git ../codeforphilly-data
+# (`--branch published` for the full laddr import.)
 CFP_DATA_REPO_PATH=../codeforphilly-data
 
 # Git remote URL to push public data commits to. Optional in dev; required in prod.

.github/workflows/ci.yml

@@ -35,5 +35,23 @@ jobs:
       - name: Test
         run: npm test
 
+      # Tests seed from apps/api/gitsheets/; fail if that copy drifts from
+      # the data repo's `empty` branch, which production configs descend from.
+      # codeforphilly-data is private: this needs a read-only deploy key in
+      # CFP_DATA_DEPLOY_KEY, and warns instead of failing without one.
+      - name: Check sheet configs against codeforphilly-data
+        env:
+          CFP_DATA_DEPLOY_KEY: ${{ secrets.CFP_DATA_DEPLOY_KEY }}
+        run: |
+          if [ -z "$CFP_DATA_DEPLOY_KEY" ]; then
+            echo "::warning title=Sheet-config drift check skipped::CFP_DATA_DEPLOY_KEY is not set, so CI can't read the private codeforphilly-data repo."
+            exit 0
+          fi
+          mkdir -p "$HOME/.ssh"
+          printf '%s\n' "$CFP_DATA_DEPLOY_KEY" > "$HOME/.ssh/id_ed25519"
+          chmod 600 "$HOME/.ssh/id_ed25519"
+          ssh-keyscan github.com >> "$HOME/.ssh/known_hosts"
+          npm run -w apps/api check:sheet-configs
+
       - name: Build
         run: npm run build

.gitignore

@@ -18,7 +18,7 @@ build/
 # it ships in the code repo so contributors can load a fake-data baseline.
 private-storage/
 
-# Dev data repo created by `npm run -w apps/api script:setup-dev-data`.
+# Local data-repo clone, if placed inside this repo.
 # Resolved relative to apps/api/ per .env.example's CFP_DATA_REPO_PATH default.
 apps/codeforphilly-data/
 codeforphilly-data/
@@ -39,3 +39,6 @@ codeforphilly-data/
 # Logs
 *.log
 npm-debug.log*
+
+# Lavish review artifacts
+.lavish/

apps/api/gitsheets/blog-posts.toml

@@ -0,0 +1,7 @@
+[gitsheet]
+root = 'blog-posts'
+path = '${{ slug }}'
+
+[gitsheet.format]
+type = 'markdown'
+body = 'body'

apps/api/gitsheets/contracts/codeforphilly.org/person-evaluations/v1.toml

@@ -0,0 +1,46 @@
+"$id" = "https://codeforphilly.org/person-evaluations/v1"
+"$schema" = "http://json-schema.org/draft-07/schema#"
+description = 'One spam verdict about one person from one evaluator. Machine evaluators (heuristic, LLM, github-probe) write these in the private codeforphilly-spam-detection repo; staff votes are written to codeforphilly-data by the site with evaluator = "human-<voterSlug>". Records are interchangeable at prune time.'
+required = ["evaluatedAt", "evaluator", "flags", "personSlug", "verdict"]
+title = "Person evaluation"
+type = "object"
+
+[properties.confidence]
+description = "0-1. LLM and probe evaluators; 1 on human votes; absent on heuristic records."
+maximum = 1
+minimum = 0
+type = "number"
+
+[properties.evaluatedAt]
+description = "ISO 8601 UTC."
+format = "date-time"
+type = "string"
+
+[properties.evaluator]
+description = "Model/run id (e.g. haiku-2026-05) or human-<voterSlug>."
+minLength = 1
+type = "string"
+
+[properties.flags]
+description = "Short reason tags."
+type = "array"
+
+[properties.flags.items]
+type = "string"
+
+[properties.personSlug]
+description = "Slug of the evaluated person."
+minLength = 1
+type = "string"
+
+[properties.reasoning]
+description = "Free-text justification. Optional on human votes."
+type = "string"
+
+[properties.score]
+description = "Heuristic evaluators only: rule points, unbounded."
+type = "integer"
+
+[properties.verdict]
+enum = ["spam", "legit", "uncertain"]
+type = "string"

apps/api/gitsheets/help-wanted-interest.toml

@@ -0,0 +1,3 @@
+[gitsheet]
+root = 'help-wanted-interest'
+path = '${{ roleId }}/${{ personSlug }}'

apps/api/gitsheets/help-wanted-roles.toml

@@ -0,0 +1,3 @@
+[gitsheet]
+root = 'help-wanted-roles'
+path = '${{ projectSlug }}/${{ id }}'

apps/api/gitsheets/people.toml

@@ -0,0 +1,3 @@
+[gitsheet]
+root = 'people'
+path = '${{ slug }}'

apps/api/gitsheets/person-evaluations.toml

@@ -0,0 +1,12 @@
+[gitsheet]
+path = "${{ personSlug }}/${{ evaluator }}"
+root = "person-evaluations"
+
+# One spam verdict per (person, evaluator). On this repo's branches the only
+# writer is the site: staff votes with evaluator = "human-<voterSlug>" (see
+# codeforphilly-ng specs/api/moderation.md). Machine evaluators (heuristic,
+# LLM) live in the private codeforphilly-spam-detection repo. The record
+# shape is the versioned contract below, which both repos and the site open
+# the sheet against; change it by adopting a new version (…/v2), never by
+# editing v1.
+implements = ["codeforphilly.org/person-evaluations/v1"]

apps/api/gitsheets/project-buzz.toml

@@ -0,0 +1,3 @@
+[gitsheet]
+root = 'project-buzz'
+path = '${{ projectSlug }}/${{ slug }}'