Echologue stores journal archives locally while routing AI tasks to cloud providers

Aris Alexis Giachnis is building Echologue, an AI voice journal that stores its permanent archive locally on the user's device while routing consented transcription and AI tasks to cloud providers including Groq and OpenRouter. The app, available on iPhone and Android, separates facts from feelings in journal entries to enable semantic retrieval without exact keyword matches. Echologue was publicly visible by June 13, 2026, and its privacy policy, last updated July 14, 2026, states that retrieval and selection of relevant entries occur locally.

Aris Alexis Giachnis https://ariregister.rik.ee/eng/company/14121026/Reformic-O%C3%9C?ref=runtimewire is building Echologue https://echologue.com/?ref=runtimewire , an AI voice journal that keeps its permanent archive on the user's device while sending consented transcription and AI tasks to outside providers. The app converts spoken or typed entries into a personal history users can question later. Echologue's product site describes it as local-first and available for iPhone and Android. Echologue was publicly visible by June 13, 2026 https://betalist.com/startups/echologue?ref=runtimewire , when BetaList featured it. At the time of reporting, Apple and Google displayed Echologue in their respective App Store https://apps.apple.com/us/app/echologue-voice-ai-journal/id6760583303?ref=runtimewire and Google Play https://play.google.com/store/apps/details?hl=en&id=com.echologue.app&ref=runtimewire storefronts. Google Play said the Android app was updated on July 21. Giachnis is trying to establish a place in a growing field of apps competing to become the memory layer for a user's private life. Giachnis is the sole shareholder, beneficial owner and board member of Reformic OU https://reformicstudio.com/?ref=runtimewire , according to Estonia's business register https://ariregister.rik.ee/eng/company/14121026/Reformic-O%C3%9C?ref=runtimewire . Reformic describes itself as a one-person studio building AI-native mobile software. A public developer profile https://diff.blog/arisAlexis/?ref=runtimewire identifies Giachnis as a full-stack developer and startup founder. Echologue is a compact expression of that solo-studio model: one specific problem, a consumer subscription and outside AI infrastructure assembled behind a mobile interface. Turning journal entries into retrieval infrastructure Echologue starts with the observation that recording a journal is easier than finding anything useful in it months later. Users dictate an unpolished voice note or type an entry. Echologue then separates the material into facts, feelings, people, places, entities and themes, and creates semantic representations that can be retrieved without an exact keyword match. A user could record a trip across several days, then ask for its highlights or how they felt during it. Echologue selects relevant entries and produces an answer grounded in those records. Other examples include asking how the user tends to feel around a particular person or which personal insights have appeared repeatedly. The separation of facts and feelings is Giachnis' clearest product choice. Many AI note products summarize everything into prose, which can blur an event with the user's interpretation of it. Echologue stores those categories separately so a question about what happened can produce different context from a question about how the user felt. The permanent archive, semantic vectors, remembered conversations and recent-state summaries live in a local database, according to Echologue's privacy policy https://echologue.com/privacy.html?ref=runtimewire . The policy says "retrieval and selection of relevant entries occur locally," and Echologue does not require an account for the local journal. Its product site https://echologue.com/?ref=runtimewire says users can export the archive as JSON or CSV. Local-first still involves cloud AI Echologue's architecture is local-first rather than fully offline. Its privacy policy, last updated July 14, 2026 https://echologue.com/privacy.html?ref=runtimewire , describes how selected data leaves the device when a user approves AI processing. That account is company-stated, with no independent audit cited by Echologue. When a user permits AI processing, voice recordings pass through what the policy calls "Echologue's Cloudflare-hosted API" to Groq for transcription. Journal text, entry context and embedding requests can pass through the same relay to OpenRouter and selected inference providers. A question posed to the journal may be transmitted with excerpts that Echologue selected locally, along with relevant profile or recent-state context, according to the policy https://echologue.com/privacy.html?ref=runtimewire . Reformic says Echologue requires affirmative permission before sending that material and blocks new third-party AI requests if permission is withdrawn in Settings. The policy https://echologue.com/privacy.html?ref=runtimewire says Reformic configures OpenRouter requests to deny data collection and use zero-data-retention endpoints where supported. It also says Reformic does not authorize providers to train general-purpose models on journal, voice or chat content. These remain Reformic's representations about services operated by outside companies. Echologue's policy https://echologue.com/privacy.html?ref=runtimewire says infrastructure and AI providers may process limited technical information, including timestamps, IP addresses, request identifiers, app or device versions, usage counts, latency, errors and security signals. It also warns that AI providers may retain limited operational metadata for security, abuse prevention, billing, legal compliance or under their published terms. On the Apple App Store listing https://apps.apple.com/us/app/echologue-voice-ai-journal/id6760583303?ref=runtimewire , Apple displays "Data Not Collected" and says Reformic OU reported that the developer does not collect data from the app. Apple notes that it has not verified the disclosure. Google Play https://play.google.com/store/apps/details?hl=en&id=com.echologue.app&ref=runtimewire displays "No data shared with third parties" and "No data collected," while saying its data-safety information was provided by the developer. Echologue's privacy policy supplies the more detailed account of temporary outside processing, including the providers and content involved. Usage limits reveal the economics At the time of reporting, Echologue's product site https://echologue.com/?ref=runtimewire described a free plan with unlimited typed entries, 20 voice notes, 15 voice minutes and 10 AI replies each month. The same page described its premium plan as including unlimited voice notes, 500 voice minutes and 400 AI replies per month, along with remembered-chat controls and custom personas. Apple's US App Store listing https://apps.apple.com/us/app/echologue-voice-ai-journal/id6760583303?ref=runtimewire listed an in-app purchase spelled "Premium Montly" at $6.99. Those quotas expose the cost boundary in Giachnis' design. Locally stored typed entries can remain unlimited, while transcription and model inference are metered. Echologue must cover recurring provider costs each time it structures an entry, generates an embedding or answers a question. The design places Reformic between users and several AI vendors, making the studio responsible for provider selection, routing rules and any future change in inference costs. A crowded market for private memory Echologue already has close competitors. Fond https://keepfond.com/?ref=runtimewire is another voice journal that says it remembers people and supports asking questions. Memex https://www.memexlab.ai/?ref=runtimewire describes an open-source, local-first AI journal for text, photos and voice, with bring-your-own-LLM support and optional Ollama workflows. Halo's website https://speaktohalo.com/?ref=runtimewire says Halo, an AI journaling companion, listens, asks a follow-up question and creates private journal entries with mood and themes. Halo says entries are private to the user's account and stored on EU servers. Lockeep https://lockeep.app/en/?ref=runtimewire says its current version performs transcription and AI organization on-device without developer-side cloud inference. Echologue's position combines managed AI processing with local permanent storage, structured fact-and-feeling extraction and natural-language recall. Giachnis is betting that users will accept temporary cloud processing when the lasting archive stays on their phones and the app explains each data flow. Echologue remains at the beginning of that test. Google Play https://play.google.com/store/apps/details?hl=en&id=com.echologue.app&ref=runtimewire displayed 1+ downloads at the time of reporting, and Apple's store https://apps.apple.com/us/app/echologue-voice-ai-journal/id6760583303?ref=runtimewire said the app had too few ratings to provide an overview. Those indicators do not reveal iOS installations and can lag actual use, though they confirm that Echologue is still seeking its initial base of regular journalers. The product becomes more useful only after someone entrusts it with weeks or months of personal history. Giachnis therefore has to earn two habits at once: frequent recording and repeated questioning. His one-person studio has built the capture, structure and retrieval loop. User trust will determine whether enough personal memory enters that loop to make Echologue worth returning to.