1version: v2
2kind: Spicepod
3name: spice-oss-docs
4
5datasets:
6  - from: github:github.com/spiceai/docs/files/trunk
7    name: spiceai.docs
8    description: Spice.ai OSS documentation and reference, from https://docs.spiceai.org
9    metadata:
10      instructions: |
11        Documents are stored in Markdown. Always provide citations.
12        When generating reference links for docs, use https://docs.spiceai.org/<docs_path> as a template
13        Exclude `website/docs` prefix from docs path and `.md` file extension.
14        Also replace `/index.md` with `/` in the path.
15      reference_base_url: https://docs.spiceai.org/<docs_path>
16    params:
17      github_client_id: ${secrets:GITHUB_CLIENT_ID}
18      github_private_key: ${secrets:GITHUB_PRIVATE_KEY}
19      github_installation_id: ${secrets:GITHUB_INSTALLATION_ID}
20      max_concurrent_requests: 1
21      include: 'website/docs/**/*.md'
22    acceleration:
23      enabled: true
24      refresh_check_interval: 4h
25      refresh_jitter_enabled: true
26      refresh_jitter_max: 30m
27    columns:
28      - name: content
29        embeddings:
30          - from: openai_embeddings
31            row_id:
32              - path
33            chunking:
34              enabled: true
35              target_chunk_size: 512
36              overlap_size: 128
37              trim_whitespace: true
38
39  - from: github:github.com/spiceai/cookbook/files/trunk
40    name: spiceai.cookbook
41    description: Spice.ai OSS cookbook recipes
42    metadata:
43      instructions: Documents are stored in Markdown. Always provide citations.
44      reference_base_url: https://github.com/spiceai/cookbook/tree/trunk/<recipe_path>
45    params:
46      github_client_id: ${secrets:GITHUB_CLIENT_ID}
47      github_private_key: ${secrets:GITHUB_PRIVATE_KEY}
48      github_installation_id: ${secrets:GITHUB_INSTALLATION_ID}
49      max_concurrent_requests: 1
50      include: '**/*.md'
51    acceleration:
52      enabled: true
53      refresh_check_interval: 4h
54      refresh_jitter_enabled: true
55      refresh_jitter_max: 30m
56    columns:
57      - name: content
58        embeddings:
59          - from: openai_embeddings
60            row_id:
61              - path
62            chunking:
63              enabled: true
64              target_chunk_size: 512
65              overlap_size: 128
66              trim_whitespace: true
67
68  - from: github:github.com/spiceai/docs/files/trunk
69    name: spiceai.releases
70    description: Spice.ai OSS release notes
71    metadata:
72      instructions: |
73        This dataset provides access to the Spice.ai OSS project release notes in Markdown format. The content is sourced from the Spice.ai OSS blog repository at https://github.com/spiceai/blog.
74      reference_base_url: https://github.com/spiceai/docs/tree/trunk/website/releases/<post_path>
75    params:
76      github_client_id: ${secrets:GITHUB_CLIENT_ID}
77      github_private_key: ${secrets:GITHUB_PRIVATE_KEY}
78      github_installation_id: ${secrets:GITHUB_INSTALLATION_ID}
79      max_concurrent_requests: 1
80      include: 'website/releases/**/*.md'
81    acceleration:
82      enabled: true
83      refresh_check_interval: 4h
84      refresh_jitter_enabled: true
85      refresh_jitter_max: 30m
86    columns:
87      - name: content
88        embeddings:
89          - from: openai_embeddings
90            row_id:
91              - path
92            chunking:
93              enabled: true
94              target_chunk_size: 512
95              overlap_size: 128
96              trim_whitespace: true
97
98  - from: github:github.com/spiceai/spiceai/issues
99    name: spiceai.issues
100    description: Spice.ai OSS issues from https://github.com/spiceai/spiceai/issues
101    params:
102      github_client_id: ${secrets:GITHUB_CLIENT_ID}
103      github_private_key: ${secrets:GITHUB_PRIVATE_KEY}
104      github_installation_id: ${secrets:GITHUB_INSTALLATION_ID}
105      max_concurrent_requests: 1
106    acceleration:
107      enabled: true
108      refresh_check_interval: 4h
109      refresh_jitter_enabled: true
110      refresh_jitter_max: 30m
111
112embeddings:
113  - from: openai
114    name: openai_embeddings
115    params:
116      openai_api_key: ${ secrets:OPENAI_API_KEY }
117
118models:
119  - name: assistant
120    from: openai:gpt-4o
121    params:
122      tools: auto
123      openai_api_key: ${secrets:OPENAI_API_KEY}
124      system_prompt: |
125        You are an AI assistant assisting engineers with the Spice.ai OSS Project.
126
127        Always strive to be accurate, concise, and helpful in your responses.
128
129        Apply instructions and reference_base_url metadata from the datasets to provide accurate and relevant information.
130
131        Prefer "docs" dataset for documentation and reference information questions.
132
133        Prefer "cookbook" recipe datasets for use cases, sample code, and configuration questions. Always include links to relevant recipes.
134
135        Use the SQL tool (sql_query) when:
136          1. The query involves precise numerical data, statistics, or aggregations.
137          2. The user asks for specific counts, sums, averages, or other calculations.
138          3. The query requires joining or comparing data from multiple related tables.
139
140        If the SQL tool returns a query, syntax, or planning error, call the `list_datasets` tool to get the available tables and continue to refine and retry the query until it succeeds. If the query fails after 5 attempts, on each subsequent run `EXPLAIN <query>` to better understand what went wrong. If it continues to fail after 10 attempts, fall back to other available tools.
141
142        When returning results from datasets, always provide citations and reference links if possible.
143
144        Use the document search tool when:
145          1. The query is about unstructured text information, such as policies, reports, or articles.
146          2. The user is looking for qualitative information or explanations.
147          3. The query requires understanding context or interpreting written content.
148
149        General guidelines:
150          1. If a query could be answered by either tool, prefer SQL for more precise, quantitative answers.
151
152  - name: docs_writer
153    from: openai:gpt-4o
154    params:
155      tools: auto
156      openai_api_key: ${secrets:OPENAI_API_KEY}
157      system_prompt: |
158        You are an AI assistant helping contributors to the Spice.ai open-source project write documentation, release notes, and blog posts.
159
160        **Write clearly, concisely, and precisely for a developer audience.**
161        Use objective third-person tense. Do not use “I” or “we” unless explicitly quoting or summarizing subjective viewpoints. Avoid vague qualifiers and revise text to maximize clarity, structure, and factual accuracy. Every sentence should reflect critical thinking and organized reasoning.
162
163        Ensure that all documentation adheres to the project's style guide and includes relevant examples where applicable.
164
165        **Be specific. Prioritize detail without being verbose.**
166        Use plain, direct language. Avoid jargon, exaggeration, and promotional tone.
167
168        **Do not use the following words or their variants:**
169        - delve
170        - seamlessly
171        - empower / empowering
172        - supercharge
173        - countless
174        - enhance / enhancing
175        - allow / allowing (Use “helps” or “enables” when appropriate.)
176
177        **Always provide citations and references with links when applicable.**
178
179        Your output should read like a well-edited technical paper: factual, sharp, and to the point.
180
181        Always output Markdown in plain text for copy.
182

1version: v2
2kind: Spicepod
3name: spice-oss-docs
4
5datasets:
6  - from: github:github.com/spiceai/docs/files/trunk
7    name: spiceai.docs
8    description: Spice.ai OSS documentation and reference, from https://docs.spiceai.org
9    metadata:
10      instructions: |
11        Documents are stored in Markdown. Always provide citations.
12        When generating reference links for docs, use https://docs.spiceai.org/<docs_path> as a template
13        Exclude `website/docs` prefix from docs path and `.md` file extension.
14        Also replace `/index.md` with `/` in the path.
15      reference_base_url: https://docs.spiceai.org/<docs_path>
16    params:
17      github_client_id: ${secrets:GITHUB_CLIENT_ID}
18      github_private_key: ${secrets:GITHUB_PRIVATE_KEY}
19      github_installation_id: ${secrets:GITHUB_INSTALLATION_ID}
20      max_concurrent_requests: 1
21      include: 'website/docs/**/*.md'
22    acceleration:
23      enabled: true
24      refresh_check_interval: 4h
25      refresh_jitter_enabled: true
26      refresh_jitter_max: 30m
27    columns:
28      - name: content
29        embeddings:
30          - from: openai_embeddings
31            row_id:
32              - path
33            chunking:
34              enabled: true
35              target_chunk_size: 512
36              overlap_size: 128
37              trim_whitespace: true
38
39  - from: github:github.com/spiceai/cookbook/files/trunk
40    name: spiceai.cookbook
41    description: Spice.ai OSS cookbook recipes
42    metadata:
43      instructions: Documents are stored in Markdown. Always provide citations.
44      reference_base_url: https://github.com/spiceai/cookbook/tree/trunk/<recipe_path>
45    params:
46      github_client_id: ${secrets:GITHUB_CLIENT_ID}
47      github_private_key: ${secrets:GITHUB_PRIVATE_KEY}
48      github_installation_id: ${secrets:GITHUB_INSTALLATION_ID}
49      max_concurrent_requests: 1
50      include: '**/*.md'
51    acceleration:
52      enabled: true
53      refresh_check_interval: 4h
54      refresh_jitter_enabled: true
55      refresh_jitter_max: 30m
56    columns:
57      - name: content
58        embeddings:
59          - from: openai_embeddings
60            row_id:
61              - path
62            chunking:
63              enabled: true
64              target_chunk_size: 512
65              overlap_size: 128
66              trim_whitespace: true
67
68  - from: github:github.com/spiceai/docs/files/trunk
69    name: spiceai.releases
70    description: Spice.ai OSS release notes
71    metadata:
72      instructions: |
73        This dataset provides access to the Spice.ai OSS project release notes in Markdown format. The content is sourced from the Spice.ai OSS blog repository at https://github.com/spiceai/blog.
74      reference_base_url: https://github.com/spiceai/docs/tree/trunk/website/releases/<post_path>
75    params:
76      github_client_id: ${secrets:GITHUB_CLIENT_ID}
77      github_private_key: ${secrets:GITHUB_PRIVATE_KEY}
78      github_installation_id: ${secrets:GITHUB_INSTALLATION_ID}
79      max_concurrent_requests: 1
80      include: 'website/releases/**/*.md'
81    acceleration:
82      enabled: true
83      refresh_check_interval: 4h
84      refresh_jitter_enabled: true
85      refresh_jitter_max: 30m
86    columns:
87      - name: content
88        embeddings:
89          - from: openai_embeddings
90            row_id:
91              - path
92            chunking:
93              enabled: true
94              target_chunk_size: 512
95              overlap_size: 128
96              trim_whitespace: true
97
98  - from: github:github.com/spiceai/spiceai/issues
99    name: spiceai.issues
100    description: Spice.ai OSS issues from https://github.com/spiceai/spiceai/issues
101    params:
102      github_client_id: ${secrets:GITHUB_CLIENT_ID}
103      github_private_key: ${secrets:GITHUB_PRIVATE_KEY}
104      github_installation_id: ${secrets:GITHUB_INSTALLATION_ID}
105      max_concurrent_requests: 1
106    acceleration:
107      enabled: true
108      refresh_check_interval: 4h
109      refresh_jitter_enabled: true
110      refresh_jitter_max: 30m
111
112embeddings:
113  - from: openai
114    name: openai_embeddings
115    params:
116      openai_api_key: ${ secrets:OPENAI_API_KEY }
117
118models:
119  - name: assistant
120    from: openai:gpt-4o
121    params:
122      tools: auto
123      openai_api_key: ${secrets:OPENAI_API_KEY}
124      system_prompt: |
125        You are an AI assistant assisting engineers with the Spice.ai OSS Project.
126
127        Always strive to be accurate, concise, and helpful in your responses.
128
129        Apply instructions and reference_base_url metadata from the datasets to provide accurate and relevant information.
130
131        Prefer "docs" dataset for documentation and reference information questions.
132
133        Prefer "cookbook" recipe datasets for use cases, sample code, and configuration questions. Always include links to relevant recipes.
134
135        Use the SQL tool (sql_query) when:
136          1. The query involves precise numerical data, statistics, or aggregations.
137          2. The user asks for specific counts, sums, averages, or other calculations.
138          3. The query requires joining or comparing data from multiple related tables.
139
140        If the SQL tool returns a query, syntax, or planning error, call the `list_datasets` tool to get the available tables and continue to refine and retry the query until it succeeds. If the query fails after 5 attempts, on each subsequent run `EXPLAIN <query>` to better understand what went wrong. If it continues to fail after 10 attempts, fall back to other available tools.
141
142        When returning results from datasets, always provide citations and reference links if possible.
143
144        Use the document search tool when:
145          1. The query is about unstructured text information, such as policies, reports, or articles.
146          2. The user is looking for qualitative information or explanations.
147          3. The query requires understanding context or interpreting written content.
148
149        General guidelines:
150          1. If a query could be answered by either tool, prefer SQL for more precise, quantitative answers.
151
152  - name: docs_writer
153    from: openai:gpt-4o
154    params:
155      tools: auto
156      openai_api_key: ${secrets:OPENAI_API_KEY}
157      system_prompt: |
158        You are an AI assistant helping contributors to the Spice.ai open-source project write documentation, release notes, and blog posts.
159
160        **Write clearly, concisely, and precisely for a developer audience.**
161        Use objective third-person tense. Do not use “I” or “we” unless explicitly quoting or summarizing subjective viewpoints. Avoid vague qualifiers and revise text to maximize clarity, structure, and factual accuracy. Every sentence should reflect critical thinking and organized reasoning.
162
163        Ensure that all documentation adheres to the project's style guide and includes relevant examples where applicable.
164
165        **Be specific. Prioritize detail without being verbose.**
166        Use plain, direct language. Avoid jargon, exaggeration, and promotional tone.
167
168        **Do not use the following words or their variants:**
169        - delve
170        - seamlessly
171        - empower / empowering
172        - supercharge
173        - countless
174        - enhance / enhancing
175        - allow / allowing (Use “helps” or “enables” when appropriate.)
176
177        **Always provide citations and references with links when applicable.**
178
179        Your output should read like a well-edited technical paper: factual, sharp, and to the point.
180
181        Always output Markdown in plain text for copy.
182