> ## Documentation Index > Fetch the complete documentation index at: https://docs.haiqu.ai/llms.txt > Use this file to discover all available pages before exploring further. # Get Haiqu Algorithm Recipe > Return a complete, runnable Python script for one algorithm. Use this tool when the user wants to run an algorithm, see its code, asks "how do I run LR-QAOA", "give me the code", "show me a noise-mitigation benchmark", or "benchmark this on a noisy device". Get the key from ``list_haiqu_algorithms`` first if the user did not give an exact one. The returned ``code`` is self-contained: it runs the algorithm across four variants (ideal baseline, noisy baseline, transpiled-on-Haiqu run, full Haiqu stack with error mitigation) on one problem instance and prints an approximation-ratio comparison. The user can run it as-is or edit the clearly marked EDIT-HERE block (graph, layers, shots, device) to change the problem. The two Haiqu variants require the user's Haiqu API key; the two baselines run with only open-source packages. Args: algorithm_key: Algorithm identifier from ``list_haiqu_algorithms`` (e.g. "lr-qaoa"). ## OpenAPI ````yaml https://api.haiqu.ai/openapi.json get /ai/qakb/algorithms/{algorithm_key}/recipe openapi: 3.1.0 info: title: Haiqu API summary: Haiqu RESTful API service. description: Cloud service providing the access to Haiqu cloud runtime contact: name: Haiqu Inc. url: https://haiqu.ai/ email: info@haiqu.ai version: 1.4.0 servers: [] security: [] paths: /ai/qakb/algorithms/{algorithm_key}/recipe: get: summary: Get Haiqu Algorithm Recipe description: >- Return a complete, runnable Python script for one algorithm. Use this tool when the user wants to run an algorithm, see its code, asks "how do I run LR-QAOA", "give me the code", "show me a noise-mitigation benchmark", or "benchmark this on a noisy device". Get the key from ``list_haiqu_algorithms`` first if the user did not give an exact one. The returned ``code`` is self-contained: it runs the algorithm across four variants (ideal baseline, noisy baseline, transpiled-on-Haiqu run, full Haiqu stack with error mitigation) on one problem instance and prints an approximation-ratio comparison. The user can run it as-is or edit the clearly marked EDIT-HERE block (graph, layers, shots, device) to change the problem. The two Haiqu variants require the user's Haiqu API key; the two baselines run with only open-source packages. Args: algorithm_key: Algorithm identifier from ``list_haiqu_algorithms`` (e.g. "lr-qaoa"). operationId: get_haiqu_algorithm_recipe parameters: - name: algorithm_key in: path required: true schema: type: string title: Algorithm Key responses: '200': description: Successful Response content: application/json: schema: $ref: '#/components/schemas/RecipeResponse' '422': description: Validation Error content: application/json: schema: $ref: '#/components/schemas/HTTPValidationError' security: - APIKeyQuery: [] - APIKeyHeader: [] components: schemas: RecipeResponse: properties: algorithm: type: string title: Algorithm description: Algorithm key the recipe is for. language: type: string title: Language description: Source language of the recipe. default: python code: type: string title: Code description: Complete, copy-pasteable Python script. Runs locally and on Haiqu. note: type: string title: Note description: How to run it + which variants need a Haiqu API key. steps: items: $ref: '#/components/schemas/RecipeStep' type: array title: Steps description: >- Ordered demo plan. The orchestrating LLM should walk these in order, rendering the indicated artifacts inline. The plan is algorithm-agnostic in shape; algorithm-specific render content comes from sandbox-side hooks the recipe defines (``compute_results``, ``format_headline``). type: object required: - algorithm - code - note title: RecipeResponse description: |- A self-contained, runnable Python script for one algorithm, plus a per-algorithm demo orchestration plan. examples: - algorithm: lr-qaoa code: | from qiskit import QuantumCircuit # ... verified internals inlined here ... MANIFEST = {'algorithm': 'lr-qaoa', 'n_qubits': 8, 'qasm': '...'} language: python note: >- Run: pip install qiskit qiskit-aer qiskit-ibm-runtime networkx numpy; python the_file.py. The haiqu_run + haiqu_full variants additionally need `pip install haiqu-sdk` and HAIQU_API_KEY; the two baselines run without them. steps: - id: show_recipe intent: User reads what is about to execute before any side effects. render_after: [] render_before: - Paste the `code` field as a fenced Python block. - Point out the EDIT-HERE region. title: Show the recipe code HTTPValidationError: properties: detail: items: $ref: '#/components/schemas/ValidationError' type: array title: Detail type: object title: HTTPValidationError RecipeStep: properties: id: type: string title: Id description: >- Stable step identifier (e.g. 'show_recipe', 'exec_recipe', 'submit_baseline_ideal'). The orchestrating LLM uses these to track progress and reference steps in subsequent turns. title: type: string title: Title description: Short human-readable title. intent: type: string title: Intent description: One-sentence explanation of why this step exists. render_before: items: type: string type: array title: Render Before description: >- Discrete items to render in chat BEFORE the step's tool call. Each item is one instruction (a code block to paste, a tool payload to render, etc.). render_after: items: type: string type: array title: Render After description: >- Discrete items to render AFTER the step's tool call returns. Include the experiment dashboard URL on every step that touches an experiment so the user can drill in mid-run. type: object required: - id - title - intent title: RecipeStep description: |- One step in the demo orchestration plan returned alongside a recipe. The orchestrating LLM walks ``RecipeResponse.steps`` in order. For each step it renders the ``render_before`` items, makes any tool call implied by the step, then renders the ``render_after`` items. Algorithm-specific content (top bitstrings vs. decoded string vs. expectation value) is not encoded here; the step instructions point the LLM at sandbox-side hooks (``compute_results``, ``format_headline``) that each recipe defines. ValidationError: properties: loc: items: anyOf: - type: string - type: integer type: array title: Location msg: type: string title: Message type: type: string title: Error Type type: object required: - loc - msg - type title: ValidationError securitySchemes: APIKeyQuery: type: apiKey in: query name: HAIQU_API_KEY APIKeyHeader: type: apiKey in: header name: authorization ````