website-brief

1# SPDX-License-Identifier: Apache-2.02# yaml-language-server: $schema=https://nika.sh/spec/v1/workflow.schema.json3#4# TEMPLATE · website-brief · crawl a site → understand it → persist the brief.5#6#   [crawl_site]────▶[brief]────▶[persist]7#   nika:fetch        infer       nika:write8#9# "Here is a URL, tell me what this company actually does." One bounded crawl,10# one typed model pass, one JSON file downstream tools can read. The output is11# a SHAPE, not prose — which is what makes it composable.12#13# Demonstrates ·14#   - `traverse:` · a bounded same-origin crawl from a single URL · one fetch15#     task, N pages, no loop to write16#   - `schema:` on `infer` · the brief is typed, so the next workflow can bind17#     `.colors` without parsing anything18#   - a JSON artifact written as a VALUE · `content:` takes one `${{ … }}`, never19#     hand-typed braces20#21# Needs · nothing to rehearse. For a REAL brief: network access and a live site22#   in `const.site_url` — then delete the `on_error:` block so a dead host is loud.23#24# Run · nika run <file> --model mock/echo    # rehearsal · green even with no network25#   (on a networked machine the crawl really fetches const.site_url — it is permitted)26nika: v127workflow:28  id: website-brief-template    # SLOT: kebab-case workflow id29  # SLOT: one honest sentence — `nika new` matches intents against these words30  description: "crawl → brief → persist"31​32# SLOT: provider/model · local · zero key. The schema below has five required33# fields, so give it a seat that holds structure — `ollama/qwen2.5:14b` is34# measured on this shape. `--model mock/echo` needs no seat at all.35model: ollama/qwen2.5:14b36​37const:38  site_url: "https://example.com"   # SLOT: the site to understand39  out_path: "./out/brief.json"      # SLOT: where the brief lands40​41⛨permits:                            # the blast radius · default-deny once present42  ⛨tools: ["nika:fetch", "nika:write"]43  ⛨net: { http: ["example.com"] }    # SLOT: the host from const.site_url. ONE entry is44                                    #   enough — `traverse:` is a same-origin BFS (+ the45                                    #   robots probe), so the crawl never leaves this host.46  ⛨fs: { write: ["./out/brief.json"] }   # SLOT: the one file · keep in step with const.out_path47                                        #   (the infer step needs no grant — it is pure compute)48​49tasks:50  crawl_site:51    ◆invoke:52      tool: "nika:fetch"53      args:54        url: "${{ const.site_url }}"55        traverse: { max_pages: 5 }  # SLOT: crawl bound · 1..=25 (robots honored)56    on_error:57      # Offline rehearsal · a literal standing in for the crawl digest, so the58      # brief step runs with no network at all. Delete once the site is real.59      recover: "REHEARSAL DIGEST · Example Co · a small tool company · plain blue and white pages · buttons, a pricing table, one logo."60​61  brief:62    with:63      crawl_site: ${{ tasks.crawl_site.output }}64    ◇infer:65      max_tokens: 120066      # SLOT: the one model job · what should the brief capture? Keep slot67      # markers out of the block below — it is prompt TEXT, not YAML, so a68      # stray comment line is sent to the model verbatim.69      prompt: |70        From this site crawl, produce a creative brief: the domain of71        activity, the dominant visual theme, the audience, the usable72        colors and image assets.73        Crawl digest · ${{ with.crawl_site }}74      schema:                       # SLOT: the typed shape downstream tasks rely on75        type: object76        additionalProperties: false77        properties:78          domain: { type: string }79          theme: { type: string }80          audience: { type: string }81          colors: { type: array, items: { type: string } }82          assets: { type: array, items: { type: string } }83        required: [domain, theme, audience, colors, assets]84​85  persist:86    with:87      brief: ${{ tasks.brief.output }}88    ◆invoke:89      tool: "nika:write"90      args:91        path: "${{ const.out_path }}"92        create_dirs: true93        # The path ends `.json`, so `content:` is ONE interpolation of a value94        # the engine serializes. Typing `{ "domain": ${{ … }} }` by hand emits95        # unquoted fields and the artifact stops being JSON.96        content: "${{ with.brief }}"97​98outputs:99  brief: ${{ tasks.brief.output }}   # SLOT: the callable contract100​